How effective is drug testing as a workplace safety strategy? A systematic review of the evidence

Full text

Accident
Analysis
and
Prevention
71
(2014)
154–165
Contents
lists
available
at
ScienceDirect
Accident
Analysis
and
Prevention
jou
rn
al
hom
ep
age:
www.elsevier.com/locate/aap
How
effective
is
drug
testing
as
a
workplace
safety
strategy?
A
systematic
review
of
the
evidence
Ken
Pidd∗,
Ann
M.
Roche1
National
Centre
for
Education
and
Training
on
Addiction,
Flinders
University,
GPO
Box
2001,
Adelaide,
SA,
5001,
Australia
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
7
November
2013
Received
in
revised
form
4
April
2014
Accepted
20
May
2014
Keywords:
Workplace
Drug
testing
Safety
Evaluation
Systematic
review
a
b
s
t
r
a
c
t
The
growing
prevalence
of
workplace
drug
testing
and
the
narrow
scope
of
previous
reviews
of
the
evidence
base
necessitate
a
comprehensive
review
of
research
concerning
the
efficacy
of
drug
testing
as
a
workplace
strategy.
A
systematic
qualitative
review
of
relevant
research
published
between
January
1990
and
January
2013
was
undertaken.
Inclusion
criteria
were
studies
that
evaluated
the
effectiveness
of
drug
testing
in
deterring
employee
drug
use
or
reducing
workplace
accident
or
injury
rates.
Methodological
adequacy
was
assessed
using
a
published
assessment
tool
specifically
designed
to
assess
the
quality
of
intervention
studies.
A
total
of
23
studies
were
reviewed
and
assessed,
six
of
which
reported
on
the
effectiveness
of
testing
in
reducing
employee
drug
use
and
17
which
reported
on
occupational
accident
or
injury
rates.
No
studies
involved
randomised
control
trials.
Only
one
study
was
assessed
as
demonstrating
strong
methodological
rigour.
That
study
found
random
alcohol
testing
reduced
fatal
accidents
in
the
transport
industry.
The
majority
of
studies
reviewed
contained
methodological
weaknesses
including;
inappropriate
study
design,
limited
sample
representativeness,
the
use
of
ecological
data
to
evaluate
individual
behaviour
change
and
failure
to
adequately
control
for
potentially
confounding
variables.
This
latter
finding
is
consistent
with
previous
reviews
and
indicates
the
evidence
base
for
the
effectiveness
of
testing
in
improving
workplace
safety
is
at
best
tenuous.
Better
dissemination
of
the
current
evidence
in
relation
to
workplace
drug
testing
is
required
to
support
evidence-informed
policy
and
practice.
There
is
also
a
pressing
need
for
more
methodologically
rigorous
research
to
evaluate
the
efficacy
and
utility
of
drug
testing.
©
2014
Elsevier
Ltd.
All
rights
reserved.
1.
Introduction
There
has
been
global
growth
in
the
use
of
workplace
drug
test-
ing
as
a
response
to
drug
related
risk
to
safety
and
productivity
(George,
2005;
Verstraete
and
Pierce,
2001;
Walsh,
2008).
The
pro-
portion
of
American
Management
Association
member
companies
that
utilised
workplace
drug
testing
grew
overall
from
21%
in
1987
to
62%
in
2004,
despite
a
19%
decline
since
the
mid
1990s
(Frone,
2013).
By
2004,
nearly
half
the
US
workforce
(48.8%)
reported
that
their
workplace
utilised
some
form
of
drug
testing
(Larson
et
al.,
2007).
While
data
on
the
prevalence
of
workplace
drug
testing
in
other
countries
is
scarce
(Pierce,
2007),
the
relatively
recent
intro-
duction
of
workplace
drug
testing
legislation
in
Australia,
Finland,
∗Corresponding
author.
Tel.:
+61
8
82017535;
fax:
+61
8
82017550.
E-mail
addresses:
ken.pidd@flinders.edu.au
(K.
Pidd),
ann.roche@flinders.edu.au
(A.M.
Roche).
1Tel.:
+61
8
82017575;
fax:
+61
8
82017550.
Ireland
and
Norway
provides
evidence
of
increasing
support
for
workplace
testing
(CASR,
1998;
Pierce,
2012).
The
main
aim
of
drug
testing
is
to
improve
workplace
safety
and
productivity.
In
the
US,
the
focus
is
on
creating
a
‘drug
free’
work-
place
by
deterring
employee
drug
use
(Frone,
2013;
MacDonald
et
al.,
1993).
The
underpinning
rationale
is
that
drug
testing
can
identify
drug
using
employees
or
job
applicants
and
these
indi-
viduals
can
be
dismissed
or
refused
employment,
thus
providing
a
deterrent
effect.
In
contrast,
the
focus
in
UK,
Canadian,
and
Australian
workplaces
is
to
improve
safety
and
productivity
by
reducing
the
incidence
or
risk
of
alcohol
or
drug
related
injuries
and
accidents
(Independent
Enquiry
into
Drug
Testing
at
Work,
2004;
Keay
et
al.,
2010;
Walker
and
Sack,
2003).
The
rationale
under-
pinning
this
aim
is
that
drug
testing
can
detect
employees
whose
performance
may
be
impaired
as
a
result
of
their
drug
use
and
thereby
any
risk
to
safety
can
be
reduced
or
eliminated
by
remov-
ing
these
employees,
either
temporarily
or
permanently,
from
the
workplace.
Despite
the
substantial
growth
in
drug
testing
as
a
workplace
strategy
and
the
apparent
logic
of
the
underlying
rationale,
the
http://dx.doi.org/10.1016/j.aap.2014.05.012
0001-4575/©
2014
Elsevier
Ltd.
All
rights
reserved.

K.
Pidd,
A.M.
Roche
/
Accident
Analysis
and
Prevention
71
(2014)
154–165
155
Addi
onal arc
les
idenfi
ed through
arcle ref
erence
lists
(n
= 4)
Arcles
idenfie
d through
datab
ase
searching
(n = 508)
Tota
l arcles
aer
dup
lica
tes
removed (n
=359)
Tota
l arcles
excluded (n
= 74
)
•Primary focus not workplace tesng (n = 62)
•
Focus
on tesn
g for
occu
paonal
exp
osu
re to chemicals (n
= 12)
Assessed arcl
es
class
ified
by
type of p
ublica
on (n = 28
5)
Eval
ua
on
(n = 27)
Reviews
(n = 17)
Des
cri
pve
(n = 13
3)
Measu
rement
(n = 42)
Policy/proced
ure
(n =35)
Lega
l/ethical
(n = 31)
Incl
uded
= 23
Excluded
= 4
Fig.
1.
Flow
chart
of
search
and
exclusion
process.
evidence
for
the
efficacy
of
testing
to
deter
employee
drug
use
or
improve
workplace
safety
appears
inconclusive.
Early
reviews
found
a
paucity
of
quality
research
and
concluded
few
definitive
conclusions
could
be
drawn
regarding
the
efficacy
of
workplace
drug
testing
(Kraus,
2001;
Macdonald
and
Wells,
1994;
Normand
et
al.,
1994).
However,
these
reviews
largely
examined
studies
undertaken
15–20
years
ago.
The
three
most
recent
reviews
of
evidence
concerning
the
effi-
cacy
of
testing
also
found
there
was
insufficient
quality
evidence
to
conclude
that
drug
testing
could
deter
use
or
prevent
occupa-
tional
injuries
(Cashman
et
al.,
2009;
Frone,
2013;
Macdonald
et
al.,
2010).
Cashman
et
al.
(2009)
undertook
a
quantitative
system-
atic
Cochrane
review
of
studies
that
examined
the
effectiveness
of
drug
testing
in
preventing
workplace
injuries
among
road
transport
drivers.
However,
this
review
was
restricted
to
a
single
industry
and
occupation
and
the
findings
may
not
generalise
to
other
indus-
tries
or
occupations.
Similarly,
Macdonald
et
al.
(2010)
focused
on
one
drug
testing
technology
(urinalysis)
to
detect
a
single
type
of
drug
(cannabis).
Employees
widely
use
cannabis,
and
urinalysis
is
a
testing
technology
routinely
used
in
the
workplace.
However,
other
illicit
drugs
are
also
relatively
prevalent
in
US
(Frone,
2006),
Euro-
pean
(Verstraete
and
Pierce,
2001)
and
Australian
(Pidd
et
al.,
2011)
workforces
and
other
types
of
testing
technologies,
such
as
saliva
analysis,
are
increasingly
common
(Walsh,
2008).
While
the
most
recent
review
(Frone,
2013)
is
more
comprehensive
and
broader
in
scope,
it
largely
summarised
the
extent
and
limitations
of
the
current
evidence
base
without
providing
a
systematic
description
of
the
methodology,
outcomes
and
quality
of
individual
studies.
Of
the
more
recent
reviews,
only
one
(Cashman
et
al.,
2009)
adopted
a
systematic
approach
to
the
selection
and
inclusion
criteria
and
the
methodological
evaluation
of
studies
reviewed.
However,
this
review
was
restricted
to
one
industry
and
occu-
pation.
While
qualitative
reviews
that
do
not
adopt
an
explicit
systematic
approach
are
informative
and
useful,
a
systematic
approach
is
essential
for
evidence
based
policy
and
practice
as
it
is
more
likely
to
identify
any
methodological
limitations
in
the
exist-
ing
evidence
and
produce
more
reliable
and
less
biased
conclusions.
To
complement
and
extend
the
previous
reviews
undertaken,
a
sys-
tematic
qualitative
literature
review
was
carried
out
to
appraise
the
effectiveness
of
workplace
drug
testing
as
a
policy
strategy
to
improve
workplace
safety
and
deter
drug
use
and
to
provide
a
comprehensive
overview
of
the
current
state
of
knowledge
in
this
area.
2.
Method
A
systematic
literature
search
was
undertaken
to
locate
and
review
research
concerning
the
effectiveness
of
workplace
drug
testing.
The
review
was
designed
to
answer
the
following
ques-
tions:
•Does
workplace
drug
testing
reduce
occupational
accident
or
injury
rates?
•Does
workplace
drug
testing
deter
employee
drug
use?

156
K.
Pidd,
A.M.
Roche
/
Accident
Analysis
and
Prevention
71
(2014)
154–165
2.1.
Search
strategy
The
CINAHL,
Medline,
PsycLit,
ProQuest
and
Web
of
Knowledge
citation
databases
were
searched
for
relevant
research
articles
published
between
January
1990
and
January
2013
using
combi-
nations
of
the
following
terms:
Work*
Alcohol
Drug
Substance
Occupation*
Employ*
Test*
Screen*
Reduc*
Impair*
Injur*
Detect*
Deter*
Death
Fatal*
Industry
A
total
of
508
articles
were
identified
(Fig.
1).
Of
these,
153
duplicates
were
removed
and
an
additional
four
papers
included
after
an
examination
of
reference
lists
provided
in
the
assessed
articles.
A
further
74
articles
were
removed
as
they
did
not
focus
on
workplace
testing
(n
=
62),
or
involved
testing
for
occupational
exposure
to
chemicals
(n
=
12).
The
remaining
285
articles
were
classified
according
to
the
main
focus
of
each
article.
The
major-
ity
(n
=
133)
were
either
descriptions
of
testing
programs,
analyses
of
psychological
and
sociological
aspects
of
testing,
or
commen-
taries.
Thirty
one
concerned
ethical
or
legal
issues,
35
policy
and
procedural
issues,
and
42
measurement
and
technical
issues.
Sev-
enteen
were
literature
reviews,
of
which
six
focused
on
evaluations
of
workplace
testing.
Of
the
remaining
27
evaluation
articles,
seven
reported
studies
that
examined
the
effectiveness
of
workplace
test-
ing
in
deterring
drug
use
and
20
reported
studies
that
examined
the
effectiveness
of
workplace
testing
in
reducing
workplace
acci-
dents
or
injuries.
One
article
that
examined
the
deterrent
effect
of
testing
(Mehay
and
Pacula,
1999)
was
excluded
as
it
focused
on
a
military
population
and
as
such
may
not
be
generalizable
to
the
civilian
workforce.
In
addition,
three
articles
that
examined
pre-
employment
screening
(Normand
et
al.,
1990;
Ryan
et
al.,
1992;
Zwerling
et
al.,
1990)
were
excluded
as
they
examined
the
relation-
ship
between
a
positive
pre-employment
testing
and
subsequent
work
outcomes,
rather
than
the
effectiveness
of
testing
in
reducing
workplace
accidents
or
injuries
per
se.
This
resulted
in
a
total
of
23
included
articles.
The
methodological
rigour
and
adequacy
of
these
studies
were
evaluated
against
a
modified
version
of
the
Effective
Public
Health
Practice
Project
(EPHPP)
qualitative
assessment
tool
for
quantita-
tive
studies
(Effective
Public
Health
Practice
Project,
1998).
The
tool
was
developed
to
assess
the
methodological
quality
of
pri-
mary
studies
in
public
health
(Thomas
et
al.,
2004)
and
is
based
on
guidelines
set
out
by
Mulrow
et
al.
(1997)
and
Jadad
et
al.
(1996).
Guidelines
provided
in
the
EPHPP
tool
dictionary
were
used
to
assess
the
methodological
adequacy
of
each
study
against
four
criteria
(selection
bias,
study
design,
confounders,
and
data
collection)
as
either
strong,
moderate,
or
weak.
Information
con-
cerning
two
other
criteria
(analysis
and
intervention
integrity)
was
summarised
(Tables
2
and
4).
As
the
review
identified
no
ran-
domised
controlled
trial
studies,
and
most
were
cross-sectional
and/or
examined
aggregated
data,
two
EPHPP
assessment
criteria
(blinding
and
withdrawal/dropouts)
were
not
included.
Based
on
the
four
assessment
criteria
utilised,
each
study
was
given
an
overall
methodological
adequacy
rating
consistent
with
EPHPP
tool
instructions.
Studies
that
obtained
at
least
two
ratings
of
strong
with
no
ratings
of
weak
for
any
of
the
assessment
criteria
were
assessed
as
methodologically
strong.
Studies
that
obtained
less
than
two
strong
ratings
but
no
more
than
one
weak
rating
for
any
of
the
assessment
criteria
were
assessed
as
methodologi-
cally
moderate.
Studies
that
obtained
two
or
more
weak
ratings
for
any
of
the
assessment
criteria
were
assessed
as
methodologically
weak.
Two
reviewers
independently
undertook
the
methodological
assessment.
For
the
majority
of
studies
assessed
(>85%)
the
review-
ers’
ratings
were
consistent.
Differences
in
assessment
ratings
were
resolved
by
discussion.
3.
Results
Of
the
23
primary
studies
included
in
the
review,
six
examined
the
efficacy
of
testing
in
deterring
drug
use
(see
Tables
1
and
2)
and
17
examined
the
efficacy
of
testing
in
reducing
occupational
injury
rates
(see
Tables
3
and
4).
All
studies
were
undertaken
in
the
USA.
3.1.
Studies
examining
deterrent
effects
All
six
of
the
deterrence
studies
indicated
that
workplace
drug
testing
was
associated
with
lower
levels
of
employee
drug
use.
One
study
utilised
a
pre/post
cohort
design
to
examine
pre-
employment
screening
among
a
sample
(N
=
987)
of
hospital
job
applicants
with
drug
test
results
as
the
outcome
variable
(Lange
et
al.,
1994).
The
remaining
five
deterrence
studies
involved
cross-
sectional
surveys
of
large
representative
samples
of
the
national
workforce
(Carpenter,
2007;
French
et
al.,
2004;
Hoffmann
and
Larison,
1999;
Larson
et
al.,
2007;
Zhang
et
al.,
1999).
All
five
of
these
cross-sectional
studies
utilised
data
collected
via
the
same
annual
national
survey
of
alcohol
and
drug
use
that
was
con-
ducted
over
different
years.
However,
while
utilising
the
same
data
source,
these
cross-sectional
studies
varied
in
the
type
of
testing
program
examined.
Two
examined
pre-employment
and
random
testing
(Hoffmann
and
Larison,
1999;
Larson
et
al.,
2007),
two
examined
pre-employment,
random,
post-accident,
and
for
cause
testing
(Carpenter,
2007;
Zhang
et
al.,
1999),
one
examined
pre-
employment,
random
and
for
cause
testing
(French
et
al.,
2004).
Only
two
studies
distinguished
between
alcohol
and
illicit
drug
testing
(Larson
et
al.,
2007;
Zhang
et
al.,
1999).
The
five
cross-sectional
studies
also
varied
in
the
self-report
measures
utilised.
Two
studies
used
a
composite
measure
of
any
illicit
drug
use
or
non-medical
use
of
pharmaceuticals
in
the
past
month
and
a
measure
of
past
month
heavy
alcohol
use,
defined
as
five
or
more
drinks
on
five
or
more
occasions
(Larson
et
al.,
2007;
Zhang
et
al.,
1999).
French
et
al.
(2004)
used
a
composite
measure
of
any
illicit
drug
use
in
the
past
year
and
current
illicit
drug
use,
defined
as
using
one
or
more
illicit
drugs
at
least
once
a
week.
Hoffmann
and
Larison
(1999)
used
a
measure
of
cocaine
and
cannabis
use
in
the
previous
3
or
more
years,
the
previous
1–3
years,
and
number
of
days
used
in
the
past
year
(1–2,
3–51
times,
51
or
more).
The
methodological
rigour
of
the
six
deterrence
studies
is
described
in
Table
2.
All
six
deterrence
studies
were
assessed
as
methodologically
weak.
Five
were
cross-sectional
studies
that
mea-
sured
drug
use
at
one
point
in
time.
Such
designs
are
inappropriate
for
examinations
of
causal
relationships.
While
Lange
et
al.
(1994)
utilised
a
more
rigorous
pre/post-test
design,
the
use
of
a
non-
random
convenience
sample,
the
lack
of
a
control
group,
and
the
use
of
organisational
level
data
to
draw
inferences
about
individual
behaviour
change
are
major
limitations.
3.2.
Studies
examining
the
effectiveness
of
testing
in
reducing
accident/injury
rates
Of
the
17
studies
that
examined
the
effectiveness
of
testing
to
reduce
workplace
accident
or
injury
rates
(Table
3),
11
reported
that
drug
testing
was
associated
with
a
reduction
in
accidents
(Brady
et
al.,
2009;
Gerber
and
Yacoubian,
2001;
Jacobson,
2003;
Miller
et
al.,
2007;
Minchin
et
al.,
2006;
Morantz
and
Mas,
2008;
Ozminkowski
et
al.,
2003;
Schofield
et
al.,
2013;
Snowden
et
al.,
2007;
Spicer
and
Miller,
2005;
Swena
and
Gaines,
1999),
three
reported
no
association
(Dell
and
Berkhout,
1998;
Kesselring
and
Pittman,
2002;
Kitterlin
and
Moreo,
2012),
and
three
reported
mixed
results
(Feinauer
and
Havlovic,
1993;
Lockwood
et
al.,
2000;
Wickizer
et
al.,
2004).
These
studies
varied
by
study
design,
the

K.
Pidd,
A.M.
Roche
/
Accident
Analysis
and
Prevention
71
(2014)
154–165
157
Table
1
Summary
of
studies
examining
testing
deterrent
effect
(from
1990
to
Jan
2013).
Study
Participants
and
setting
Test
type
Design
Measures
Outcome
Lange
et
al.
(1994)
USA
987
hospital
job
applicants
(21%
male;
77%
<35
yrs
old)
Pre-employment
Pre/post
Urine
test
results
T1
(1989)
and
T2
(1991)
Positive
test
rate
declined
significantly
from
10.8%
(T1)
to
5.8%
(T2)
Hoffmann
and
Larison
(1999)
USA
9097
employed
respondents
to
1994
NHSDA
(54%
male;
Mage =
39
yrs)
Pre-employment
and
random
Cross-sectional
Frequency
of
cannabis
and
cocaine
use
(self-report)
Frequent
drug
use
significantly
less
likely
among
employees
of
companies
that
test
than
employees
of
companies
that
do
not
test
Zhang
et
al.
(1999)
USA 15,012
employed
respondents
(aged
18–49
yrs)
to
the
1994
(58.9%
male)
and
1997
NHSDA
(59.5%
male)
Pre-employment
random,
for
cause
and
post-accident
(alcohol
and
drug)
Cross-sectional
Past
month
any
drug
use
(self-report)
Past
month
heavy
alcohol
use
(self-report)
Compared
to
1994
survey
respondents
employed
in
non-testing
companies,
drug
use
was
significantly
less
likely
for
1994
survey
respondents
employed
in
companies
that
conducted
random
or
pre-employment
tests.
No
significant
differences
in
heavy
alcohol
use
by
drug
testing
status
were
observed
No
significant
difference
in
drug
use
or
heavy
alcohol
use
by
employer
testing
status
was
observed
for
1997
survey
respondents
French
et
al.
(2004)
USA
15,400
employed
respondents
to
the
1997
and
1998
NHSDA
(46.5%
male;
Mage =
36.4
yrs)
Pre-employment
for
cause
and
random
Cross-sectional
Past
week
any
drug
use
(self-report)
Past
year
any
drug
use
(self-report)
Past
year
and
past
week
drug
use
was
significantly
less
likely
for
employees
of
companies
with
random,
for-cause
or
pre-employment
drug
testing
compared
to
employees
of
companies
that
did
not
test
Carpenter
(2007)
USA 57,397
employed
2000
and
2001
NHSDA
and
2002
NSDUH
respondents
(58%
male;
40%
<35
yrs
old)
Pre-employment
random,
for
cause
and
post-accident
Cross-sectional
Past
month
cannabis
use
(self-report)
Cannabis
use
significantly
less
likely
among
employees
of
testing
companies,
compared
to
employees
of
non-testing
companies
A
similar
significant
effect
was
observed
for
employees
of
companies
that
did
not
test
but
provided
either
drug
education,
a
written
policy,
or
an
EAP
Larson
et
al.
(2007)
USA
73,325
employed
2002,
2003,
and
2004
NSDUH
respondents
(gender
%
and
Mage not
reported)
Pre-employment
and
random
(alcohol
and
drug)
Cross-sectional
Past
month
any
drug
use
(self-report)
Past
month
heavy
alcohol
use
(self-report)
Past
month
drug
use
significantly
less
for
employees
of
testing
companies
compared
to
employees
of
non-testing
companies
No
significant
difference
in
past
month
heavy
alcohol
use
between
employees
of
testing
and
non-testing
companies
NHSDA
=
US
National
Household
Survey
on
Drug
Abuse;
NSDUH
=
US
National
Survey
on
Drug
Use
and
Health.
The
NHSDA
and
the
NSDUH
are
the
same
annual
survey.
In
2002
the
name
of
the
National
Household
Survey
on
Drug
Abuse
was
changed
to
the
National
Survey
on
Drug
Use
and
Health.
type
of
testing
program
examined,
outcome
measures,
and
meth-
odological
rigour.
Study
designs
included
time
series,
cross-sectional,
pre/post,
and
matched
pairs.
Seven
studies
utilised
a
time
series
design
involving
610
metal
foundry
employees
(Dell
and
Berkhout,
1998),
219
hotel
employees
(Lockwood
et
al.,
2000),
1791
manufac-
turing
employees
(Ozminkowski
et
al.,
2003),
256
state
based
companies
(Wickizer
et
al.,
2004),
employees
of
five
transport
com-
panies
(Spicer
and
Miller,
2005),
26,000
rail
transport
company
employees
(Miller
et
al.,
2007),
and
the
national
truck
driving
workforce
(Swena
and
Gaines,
1999;
Snowden
et
al.,
2007).
Six
studies
were
cross-sectional
surveys
of
69
(Gerber
and
Yacoubian,
2001)
and
34
(Minchin
et
al.,
2006)
construction
company
repre-
sentatives,
110
restaurant
management
representatives
(Kitterlin
and
Moreo,
2012),
a
representative
sample
of
the
national
work-
force
(Kesselring
and
Pittman,
2002),
the
national
commercial
truck
driving
workforce
(Jacobson,
2003)
and
one
was
a
cross-sectional
analysis
of
workers’
compensation
data
for
1360
construction

158
K.
Pidd,
A.M.
Roche
/
Accident
Analysis
and
Prevention
71
(2014)
154–165
Table
2
Summary
of
the
methodological
rigour
of
studies
examining
testing
deterrent
effect
(from
1990
to
Jan
2013).
Study
Control
for
selection
bias
Study
design
Control
for
confounding
Data
collection
Analysis
Intervention
integrity
Overall
rating
Lange
et
al.
(1994)
Weak
Moderate
Weak
Moderate
No
sample
size
calculation
Unit
of
analysis:
organisation
Post-intervention
sample
not
the
same
as
pre-intervention
sample
Weak
Hoffmann
and
Larison
(1999)
Strong
Weak
Weak
Moderate
Unit
of
analysis:
individual
Types
of
testing
programs
varied
among
survey
respondents
Frequency
of
testing
and
proportion
of
employees
tested
not
reported
Weak
Zhang
et
al.
(1999)
Strong
Weak
Weak
Moderate
Unit
of
analysis:
individual
Types
of
testing
programs
varied
among
survey
respondents
Frequency
of
testing
and
proportion
of
employees
tested
not
reported
Weak
French
et
al.
(2004)
Strong
Weak
Weak
Moderate
Unit
of
analysis:
individual
Types
of
testing
programs
varied
among
survey
respondents
Frequency
of
testing
and
proportion
of
employees
tested
not
reported
Weak
Carpenter
(2007)
Strong
Weak
Weak
Moderate
Unit
of
analysis:
individual
Types
of
testing
programs
varied
among
survey
respondents
Frequency
of
testing
and
proportion
of
employees
tested
not
reported
Weak
Larson
et
al.
(2007)
Strong
Weak
Weak
Moderate
Unit
of
analysis:
individual
Types
of
testing
programs
varied
among
survey
respondents
Frequency
of
testing
and
proportion
of
employees
tested
not
reported
Weak
companies
(Schofield
et
al.,
2013).
Two
studies
were
pre/post
designs
with
employees
of
a
large
retail
chain
(Morantz
and
Mas,
2008)
and
48
manufacturing
and
service
industry
workplaces
(Feinauer
and
Havlovic,
1993).
One
study
utilised
a
matched
pairs
design
that
examined
the
national
trucking
industry
workforce
(Brady
et
al.,
2009).
Four
studies
examined
pre-employment,
random,
for
cause,
and
post-accident
testing
(Jacobson,
2003;
Minchin
et
al.,
2006;
Ozminkowski
et
al.,
2003;
Schofield
et
al.,
2013).
Five
studies
examined
random
testing,
two
of
which
examined
random
alco-
hol
and
random
drug
testing
(Miller
et
al.,
2007;
Spicer
and
Miller,
2005),
while
two
examined
only
random
alcohol
testing
(Brady
et
al.,
2009;
Snowden
et
al.,
2007)
and
one
examined
only
ran-
dom
drug
testing
(Swena
and
Gaines,
1999).
Two
studies
examined
pre-employment
testing
(Dell
and
Berkhout,
1998;
Kitterlin
and
Moreo,
2012).
One
study
examined
pre-employment
and
random
testing
(Lockwood
et
al.,
2000),
one
examined
pre-employment
and
for
cause
testing
(Feinauer
and
Havlovic,
1993)
and
one
exam-
ined
post-accident
testing
(Morantz
and
Mas,
2008).
The
remaining
three
studies
did
not
report
the
type
of
testing
examined
(Gerber
and
Yacoubian,
2001;
Kesselring
and
Pittman,
2002;
Wickizer
et
al.,
2004).
Outcome
measures
of
four
studies
were
national
road
accident
fatality
rates
among
trucking
industry
employees
(Brady
et
al.,
2009;
Jacobson,
2003;
Snowden
et
al.,
2007;
Swena
and
Gaines,
1999).
The
outcome
measures
in
the
remaining
13
studies
were
all
indicators
of
workplace
fatal
and
non-fatal
accident/injury
rates.
Two
of
these
studies
relied
on
self-report
data
(Kitterlin
and
Moreo,
2012;
Minchin
et
al.,
2006),
while
11
accessed
reportable
injury
data
records.
Of
these
11
studies,
one
measured
accident/injury
rates
at
a
state
level
(Kesselring
and
Pittman,
2002),
one
at
a
geo-
graphical
district
level
involving
more
than
one
workplace
from
the
same
organisation
(Morantz
and
Mas,
2008),
one
at
the
organisa-
tional
and
national
level
(Miller
et
al.,
2007),
while
the
remaining
studies
measured
injury
rates
at
the
organisational
level.
In
terms
of
methodological
rigour,
eight
studies
(47.1%)
were
assessed
as
methodologically
weak,
eight
were
assessed
as
demon-
strating
moderate
methodological
rigour,
and
one
was
assessed
and
methodologically
strong.
Five
of
the
methodologically
weak
studies
measured
injury
rates
at
a
single
point
in
time
using
a
cross-sectional
survey
(Gerber
and
Yacoubian,
2001;
Kesselring
and
Pittman,
2002;
Kitterlin
and
Moreo,
2012;
Minchin
et
al.,
2006;
Schofield
et
al.,
2013);
a
design
that
is
inappropriate
for
exam-
inations
of
causal
relationships.
The
other
three
studies
adopted
more
rigorous
designs,
but
were
flawed
by
potential
selection
bias
and
failure
to
control
for
confounding
variables
(Dell
and
Berkhout,
1998;
Feinauer
and
Havlovic,
1993;
Lockwood
et
al.,
2000).
Five
of
the
eight
studies
that
were
assessed
as
demonstrating
moderate
methodological
adequacy
found
random
alcohol
and
drug
(Miller
et
al.,
2007;
Spicer
and
Miller,
2005),
random
drug
(Swena
and
Gaines,
1999;
Jacobson,
2003),
random
alcohol
(Snowden
et
al.,
2007),
and
post-accident
(Morantz
and
Mas,
2008)
testing
pro-
grams
were
associated
with
reductions
in
injury
rates.
Of
the
remaining
two
studies
one
that
failed
to
report
the
type
testing
program
(Wickizer
et
al.,
2004)
also
found
testing
was
associated
with
a
reduction
in
injury
rates,
while
one
that
examined
a
com-
prehensive
testing
program
(Ozminkowski
et
al.,
2003)
reported
a
non-significant
reduction
in
injury
rates.
However,
the
methodo-
logical
rigour
of
these
seven
studies
was
limited
by
the
failure
to
adequately
account
for
potential
confounding
variables
that
may
have
influenced
any
observed
effect.
Only
one
study
was
assessed
as
demonstrating
strong
method-
ological
rigour.
Brady
et
al.
(2009)
adopted
a
quasi-experimental
design
that
compared
the
prevalence
of
alcohol
involvement
in

K.
Pidd,
A.M.
Roche
/
Accident
Analysis
and
Prevention
71
(2014)
154–165
159
Table
3
Summary
of
studies
evaluating
the
effectiveness
of
testing
in
reducing
workplace
accidents/injuries
(from
1990
to
Jan
2013).
Study
Participants
and
setting
Test
type
Design
Measures
Outcome
Feinauer
and
Havlovic
(1993)
USA
48
Wisconsin
manufacturing
and
service
industry
companies
(employee
sample
size,
gender
and
age
not
reported)
Pre-
employment
post-accident
and
for
cause
Cross-
sectional/time
series
Aggregated
accident
rate
(5
years
period)
No
significant
difference
in
accident
rates
between
12
testing
and
36
non-testing
companies
Pooled
time
series
regression
analysis
of
injury
data
for
the
12
testing
companies
indicated
post-accident
testing
was
associated
with
a
significant
reduction
in
accident
rates
Dell
and
Berkhout
(1998)
USA
610
metal
foundry
employees
(gender
and
age
not
reported)
Pre-
employment
Interrupted
time
series
Monthly
accident
data
(1980–1994)
Introduction
of
testing
not
significantly
associated
with
a
reduction
in
monthly
accident
rate
Swena
and
Gaines
(1999)
Interstate
commercial
truck
drivers
(sample
size,
gender
and
age
not
reported)
Random
Interrupted
time
series
Fatality
rate
per
100
mile
travelled
(1988–1990)
A
significant
reduction
in
road
fatality
rates
was
observed
for
2
years
post
testing
implementation
Lockwood
et
al.
(2000)
USA
219
employees
of
three
Florida
hotels
(gender
and
age
not
reported)
Pre-
employment
and
random
Interrupted
time
series
Monthly
accident
rate
(1992–1996)
No
significant
reduction
in
accident
rates
for
the
two
hotels
that
introduced
pre-employment
testing,
but
a
significant
reduction
in
accident
rates
was
observed
for
the
one
hotel
that
introduced
both
random
and
pre-employment
testing
Gerber
and
Yacoubian
(2001)
USA
Management
representatives
of
69
construction
companies
(employee
sample
size,
gender
and
age
not
reported)
Not
reported
Cross-sectional
Occupational
accident/illness
incident
rate
MODarating
Comparison
of
drug
testing
and
non-drug
testing
companies
indicated
drug
testing
was
associated
with
a
significantly
lower
occupational
accident/illness
rates
and
a
significantly
lower
MOD
rating
Kesselring
and
Pittman
(2002)
USA
National
workforce
(gender
and
age
not
reported)
Not
reported Cross-sectional
Occupational
accident/illness
incident
rate
No
significant
difference
in
occupational
accident/illness
rates
between
states
according
to
state
based
workplace
drug
testing
legislation
category
(restrictive,
supportive,
neutral,
or
none)
Jacobson
(2003)
USA
National
commercial
truck
driving
workforce
(gender
and
age
not
reported)
Pre-
employment
for
cause,
post-accident,
random
Cross-
sectional/time
series
Aggregated
state
fatal
truck
accident
rate
(1983–1998)
A
significant
reduction
in
road
fatality
rates
(9–10%)
was
associated
with
the
introduction
of
testing
Ozminkowski
et
al.
(2003)
USA
1791
manufacturing
company
employees
(gender
and
age
not
reported)
Pre-
employment
for
cause,
post-accident,
random
Interrupted
time
series
Monthly
injury
rate
(1996–1999)
Drug
testing
rate
associated
with
a
non-significant
(p
=
.053)
reduction
in
injury
rates
Wickizer
et
al.
(2004)
USA
256
Washington
State
companies
(employee
sample
size,
gender
and
age
not
reported)
Not
reported
Pre/post
with
comparison
group
Injury
incidence
rates
(1994–2000)
Drug
free
program
associated
with
significant
reduction
in
reported
injury
rates
for
three
of
nine
industries
examined
(construction,
manufacturing
and
services)
and
in
lost
time
(>4
days)
injury
rates
for
two
industries
(construction
and
services)
Spicer
and
Miller
(2005)
USAb
Approx.
125,000
employees
of
five
transport
industry
companies
(gender
and
age
not
reported)
Random
alcohol
and
drug
Interrupted
time
series
Monthly
injury
rate
(1983–1996)
Industry
wide
implementation
of
both
random
drug
and
alcohol
testing
was
associated
with
a
significant
decline
in
injury
risk
A
peer
intervention
program
(Peer
Care)
introduced
into
one
of
the
five
companies
was
also
associated
with
a
significant
decline
in
injury
risk
prior
to
the
introduction
of
random
drug
or
alcohol
testing
Minchin
et
al.
(2006)
USA
34
Florida
based
construction
companies
(gender
and
age
not
reported)
Pre-
employment
for
cause,
post-accident,
and
random
Cross-sectional
Accident
rates
(self-report
by
company
rep-
resentatives)
Survey
respondents
employed
in
29
companies
that
tested
reported
a
10–60%
reduction
in
accident
rates
post
testing
implementation
Miller
et
al.
(2007)
USAb
26,000
employees
of
a
rail
transport
company
(gender
and
age
not
reported)
Random
alcohol
and
drug
Interrupted
time
series
with
comparison
group
Monthly
injury
rate
(1983–1999)
The
combination
of
a
peer-based
program
(Peer
Care)
and
drug
testing
was
associated
with
an
approximate
one-third
reduction
in
injury
rates.
Alcohol
testing
and
related
Peer
Care
enhancements
were
associated
with
a
further
17%
reduction
in
injuries