Prevalence and distribution of illicit drug use in the workforce and in the workplace: Findings and implications from a U.S. National survey.

Abstract

This U.S. national study explored the overall prevalence, frequency, and distribution of illicit drug use in the workforce and in the workplace during the preceding 12 months. Illicit drug use in the workforce involved an estimated 14.1% of employed adults (17.7 million workers). Illicit drug use in the workplace involved an estimated 3.1% of employed adults (3.9 million workers). Illicit drug use in the workforce and in the workplace is not distributed uniformly in the employed population. At-risk, though circumscribed, segments of the U.S. workforce were identified with prevalence rates up to 55.8% for any use of illicit drugs and up to 28.0% for illicit drug use in the workplace. The implications of these data for future theoretical research and for management policy and practice are discussed.

Full text

Prevalence and Distribution of Illicit Drug Use in the Workforce and in the
Workplace: Findings and Implications From a U.S. National Survey
Michael R. Frone
University at Buffalo, State University of New York
This U.S. national study explored the overall prevalence, frequency, and distribution of illicit drug use
in the workforce and in the workplace during the preceding 12 months. Illicit drug use in the workforce
involved an estimated 14.1% of employed adults (17.7 million workers). Illicit drug use in the workplace
involved an estimated 3.1% of employed adults (3.9 million workers). Illicit drug use in the workforce
and in the workplace is not distributed uniformly in the employed population. At-risk, though circum-
scribed, segments of the U.S. workforce were identified with prevalence rates up to 55.8% for any use
of illicit drugs and up to 28.0% for illicit drug use in the workplace. The implications of these data for
future theoretical research and for management policy and practice are discussed.
Keywords: workforce, workplace, prevalence, substance use, illicit drugs
The use of illicit drugs by employed adults represents an im-
portant social policy issue because it may undermine employee
health, productivity, and safety (e.g., Frone, 2004; Roman & Blum,
1995; Sindelar, 1998). Such effects may further interfere with an
employer’s ability to compete effectively in an increasingly com-
petitive domestic and global economic environment. One sign of
the presumed importance of dealing with illicit drug use is the
billions of dollars spent by the U.S. government each year on law
enforcement, corrections, prevention, treatment, and research. A
second sign is the passage of the Drug-Free Workplace Act of
1988. A third sign is the growing number of work organizations
that have implemented formal drug use policies, drug testing
programs, and employee assistance programs. But what does re-
search suggest regarding the scope of the potential illicit drug use
problem for U.S. employers? Although national data exist regard-
ing the overall prevalence of illicit drug use in the workforce,
much less published information exists on the frequency of drug
use and the distribution of use in the workforce. Further, little
published literature exists on the prevalence, frequency, and dis-
tribution of illicit drug use in the workplace (i.e., during the
workday). Unfortunately, overall prevalence rates for illicit drug
use in the workforce provide insufficient information for manag-
ers, policymakers, and researchers regarding the scope of the
potential problem in the workforce and in the workplace.
Although often viewed as atheoretical and descriptive, data on
the prevalence, frequency, and distribution of workforce and work-
place drug use can be very useful to practicing managers, policy-
makers, and researchers (Schaufeli, 2004; Wittchen, 2004).
Schaufeli (2004) identified descriptive epidemiological studies
among the five types of research that underlie the development of
the field of occupational health psychology. In addition, Harris
(2004) pointed out that exploring prevalence is one way to gauge
whether illicit drug use is a problem for organizations. Thus, as a
first step in theoretical research and policy discussions on social
problems, it is valuable to have high-quality data on the extent and
distribution of the social problem in the population of interest.
Further, a lack of a clear understanding of the prevalence, fre-
quency, and distribution of illicit drug use and impairment in the
workforce and workplace may undermine the development of
strong theoretically based research exploring the antecedents and
outcomes of illicit drug use among employees, and it interferes
with the development of defensible workplace policies and inter-
ventions designed to change illicit drug use among employees.
Therefore, in this article, I will address four general goals. The
first goal is to provide some definitions and highlight some issues
that have been overlooked by researchers and policymakers. The
second goal is to review past research on the prevalence of illicit
drug use in the workforce and in the workplace. The third goal is
to explore in detail the prevalence, frequency, and distribution of
illicit drug use in the workforce and in the workplace, with an eye
toward identifying those segments of the U.S. workforce at great-
est risk. The final goal is to discuss the specific implications of the
data presented for future theoretically based research and for
management practice and policy.
Definitions and Issues
To better understand the issue of illicit drug use in the popula-
tion of employed adults and its implications for employers, one
needs to consider two critical and overlooked issues (Frone, 2004).
The first issue is the distinction between drug use and impairment.
Measures of illicit drug use reflect the mere use of a substance—
that is, the prevalence or frequency of using a substance over some
fixed period of time or the quantity of a substance consumed on a
typical occasion of use. A major dimension of illicit drug impair-
ment is intoxication. In the medical and pharmacological litera-
tures, intoxication simply refers to a state of being poisoned by
some substance. Therefore, illicit drug intoxication refers to re-
Data collection was supported by National Institute on Alcohol Abuse
and Alcoholism Grant R01-AA12412 to Michael R. Frone.
Correspondence concerning this article should be addressed to Michael
R. Frone, Research Institute on Addictions, University at Buffalo, State
University of New York, 1021 Main Street, Buffalo, NY 14203. E-mail:
frone@ria.buffalo.edu
Journal of Applied Psychology Copyright 2006 by the American Psychological Association
2006, Vol. 91, No. 4, 856 – 869 0021-9010/06/$12.00 DOI: 10.1037/0021-9010.91.4.856
856
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

versible central nervous system impairment due to the direct
pharmacological action of a substance resulting in various behav-
ioral, cognitive, and affective changes (e.g., American Psychiatric
Association, 1994). The distinction between drug use and impair-
ment is important because of the often-ignored issue of acquired
tolerance to the pharmacological effects of a drug. There are
several general forms of tolerance, such as pharmacokinetic toler-
ance, pharmacodynamic tolerance, and behavioral or learned tol-
erance (see Frone, 2004, for a detailed discussion of this issue).
Despite the several forms of tolerance, the general issue is that
with repeated exposure, a process of physiological adaptation takes
place whereby the central nervous system adjusts to the constant
presence of the drug, or a process of behavioral adaptation takes
place whereby compensatory behaviors mitigate the physiological
impact of a drug. Thus, a person builds up resistance to the effects
of a specific dose of a given drug. This means that simply knowing
the overall prevalence of illicit drug use in the population of
employed adults provides little information regarding the preva-
lence of illicit drug impairment in the workforce and in the
workplace. When one considers illicit drug use and impairment, it
is the latter that may be the more proximal cause of poor work-
place productivity outcomes (Frone, 2004).
The second issue that has been overlooked is the context of
employee drug use and impairment. Researchers have typically
only assessed employees’ overall use of and impairment from
illicit drugs. Overall illicit drug use represents the consumption of
illicit drugs across all contexts of use. Overall illicit drug impair-
ment represents impairment (i.e., level of intoxication) due to drug
use across all contexts of use. Thus, past research has primarily
explored illicit drug use and impairment in the workforce, which
largely reflects use and impairment away from work and outside
an employed individual’s normal work hours. In contrast, past
research has not generally focused on drug use in the workplace
even though such information is important to employers and
policymakers. Illicit drug use in the workplace represents the
consumption of drugs at times that occur just before or during
one’s formal work hours (Ames, Grube, & Moore, 1997; Frone,
2004). Specifically, workplace illicit drug use refers to the con-
sumption of illicit drugs (a) within 2 hr of starting one’s work shift,
(b) during a lunch break, (c) during other work breaks, and (d)
while performing one’s job. And workplace illicit drug impairment
represents impairment (i.e., levels of intoxication) due to illicit
drug use experienced during work hours.
Illicit Drug Use in the Workforce
Several reports have used representative, national data to exam-
ine the prevalence of illicit drug use in the U.S. workforce (e.g.,
Hoffman, Brittingham, & Larison, 1996; Hoffman, Larison, &
Sanderson, 1997). For example, data taken from the 1993 National
Household Survey on Drug Abuse show that among employed
adults (ages 18 – 49) who work full time, approximately 15%
reported using illicit drugs at least once during the past year (e.g.,
Hoffman et al., 1996). Nonetheless, these reports have failed to
address adequately three issues. First, the prevalence of workforce
drug use has been explored across a limited set of demographic
characteristics, and the data presented represent bivariate relations.
Thus, past research has failed to consider some potentially impor-
tant demographic characteristics, and confounding among the de-
mographic variables (e.g., between gender and occupation) has not
been addressed. Second, little information is provided on the
frequency of drug use in the workforce. Finally, no information is
provided on whether any impairment resulted from use.
Illicit Drug Use in the Workplace
The prevalence rate reported earlier for illicit drug use in the
workforce may suggest that illicit drugs are being used in the
workplace or that workers may be arriving impaired at work.
However, Hoffman et al. (1997) clearly warned that such data are
not a good proxy for illicit drug use and impairment in the
workplace because they fail to account for the circumstances or
timing of use. Research, therefore, needs to explicitly assess illicit
drug use and impairment in the workplace. Only two studies of
workplace drug use have used U.S. national samples. Using data
from the 1984 National Longitudinal Survey of Youth (N ⫽
12,069), Gleason, Veum, and Pergamit (1991) reported that 9.5%
of men and 4.2% of women used an illicit drug on the job during
the past 12 months. Unpublished data from the 1991 follow-up
surveys of the annual Monitoring the Future Study showed that
marijuana was used at work by 5% of men and 1% of women, and
cocaine, amphetamines, and tranquilizers were each used by less
than 1% of men and women (Normand, Lempert, & O’Brien,
1994). Although Gleason et al. (1991) used a 1984 national sam-
ple, these data are now 20 years old. Also, because of the sample’s
narrow age range (19 –27 years old), the reported prevalence
estimates cannot be generalized to the overall U.S. workforce, and
workplace prevalence cannot be explored in other segments of the
workforce. The Monitoring the Future Study data are limited by
the fact that they are based on follow-up surveys of a national
sample of high school seniors. Because the original sample did not
include high school dropouts and the follow-up surveys lost re-
spondents to attrition, the data from this study are actually not
representative of the U.S. workforce even for the restricted age
range (19 –28 years old) covered by the sample. Moreover, these
data are more than 14 years old. Besides these two national
samples, other studies have used at times small, convenience
samples drawn from specific communities or specific companies
or occupations (Newcomb, 1994). For example, Newcomb (1989,
cited in Newcomb, 1994) reported that 27% of men (N ⫽ 154) and
13% of women (N ⫽ 391) from a community sample of young
workers were “drunk, high, or stoned” at work from the use of an
illicit drug. This study by Newcomb (1989, cited in Newcomb,
1994) is unique in that it was the only study to ask about illicit drug
impairment at work.
In addition to sampling limitations, several other problems exist
in past research on illicit drug use in the workplace. In a report for
the National Research Council, Normand et al. (1994) concluded
that the meaning of workplace drug use is not well defined in the
few studies that have explored this issue. Often questions simply
ask about the use of some drug “at work” without an explicit
definition of what this term means or includes. Also, no single
study has asked about both workplace drug use and impairment.
Finally, no data exist on the frequency of workplace illicit drug
use, and very little data exist on the distribution of workplace illicit
drug use and impairment across key demographic and occupa-
tional characteristics. It is not surprising then that Normand et al.
recommended that a national study of the prevalence of workplace
857
ILLICIT DRUG USE IN THE WORKFORCE AND WORKPLACE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

drug use be conducted. In effect, Normand et al. concluded that
relevant stakeholders (e.g., researchers, managers, and policymak-
ers) do not have sufficient credible scientific data regarding illicit
drug use in the workplace.
The Present Study
The present study will examine the prevalence and distribution
of illicit drug use and impairment in the workforce in more detail
than has been done in past research. More important, this study
answers Normand et al.’s (1994) call for a national epidemiologic
study of illicit drug use in the workplace. For illicit drug use in the
workforce, I will report the frequency of use and the overall
proportion of the workforce who have engaged in the illicit use of
marijuana, cocaine, and prescription psychotherapeutic drugs (i.e.,
sedatives, tranquilizers, stimulants, and analgesics) during the pre-
ceding 12 months. To look at impairment in the workforce, I will
report the frequency and prevalence of taking enough of each drug
to get high or stoned. For illicit drug use in the workplace, I will
present data on the frequency and overall prevalence of using the
same drugs before reporting to work and at several points during
the workday. Also, to look at impairment in the workplace, I will
report the frequency and overall prevalence of being at work high
on or under the influence of these drugs. In addition to these
prevalence estimates, I will provide estimates of the number of
employees (i.e., population totals) involved in the U.S. workforce.
The distribution of workforce and workplace illicit drug use and
impairment will be explored by general demographic and occupa-
tional demographic characteristics. The specific variables used
were largely selected because they have been used in prior studies,
though some potentially important variables that have not been
used in prior research are included (e.g., number of financial
dependents, work shift, and union membership). Both bivariate
and multivariate relations will be presented. The results from the
multivariate analyses will be used to identify those subgroups at
greatest risk for illicit drug use in the workforce and in the
workplace.
Method
Study Design
The 2,829 study participants took part in the National Survey of Work-
place Health and Safety. This telephone survey was designed to explore a
wide range of issues related to employee health and safety. The population
from which the study participants were sampled comprised noninstitution-
alized adults aged 18– 65 who were employed in the civilian labor force
and residing in households in the 48 contiguous United States and the
District of Columbia. A list-assisted, two-stratum truncated design (e.g.,
Levy & Lemeshow, 1999; Tucker, Lepkowski, & Piekarski, 2002) was
used to identify the sampling frame of telephone numbers. The numbers in
the sampling frame were then stratified by county, and the actual sample of
telephone numbers was selected from the sampling frame using systematic
sampling. Data were collected by 19 extensively trained interviewers using
computer-assisted telephone interviewing stations from January 2002 to
June 2003. All interviewers received 4 weeks of extensive training on
general and study-specific issues. The issues included the role of surveys,
the role of interviewers in the survey process, general interviewing tech-
niques and etiquette, handling common questions and techniques to secure
an interview, study-specific interviewing and probing issues, and super-
vised instruction and hands-on practice with the computer-assisted tele-
phone interviewing equipment and the actual interview. Each telephone
number was called up to 20 times to screen for a working household
number, to determine eligibility of the household, and to select an eligible
respondent. For telephone numbers associated with an eligible individual
who was selected to participate, each was called up to an additional 20
times in an effort to secure an interview with the selected respondent.
Within a household with more than one eligible individual, the most recent
birthday method was used to select at random one individual for partici-
pation in the study (e.g., Potthoff, 1994). Of all selected eligible individ-
uals, 57% participated in the study. Before being interviewed, all partici-
pants gave informed consent. Also, each participant was informed that a
Certificate of Confidentiality was obtained from the National Institutes of
Health in order to assure the confidentiality of responses and the privacy of
study participants. This certificate assures confidentiality and privacy to
study participants by allowing the investigator and others who have access
to research records to refuse to disclose identifying information on research
participants in any civil, criminal, administrative, legislative, or other
proceeding, whether at the federal, state, or local level. On average, the
interview lasted 45 min, and participants were paid $25 for their time.
Of the 2,829 study participants, the present analyses were restricted to
the 2,806 workers who had complete data on all of the variables used in this
report. In general, the percentage of missing data on the variables included
in this report was low (M ⫽ 0.3%, range ⫽ 0%– 0.7%). Therefore, listwise
deletion eliminated only 23 cases or 0.81% of the sample.
Sampling Weights
For all analyses, the data were weighted according to standard proce-
dures for sample survey data (e.g., Korn & Graubard, 1999; Levy &
Lemeshow, 1999) so the results can be generalized to the U.S. workforce.
Several general steps went into the computation of the sampling weights.
In Step 1, the initial base weight for each interviewee was a function of the
selection probability for the reached telephone number, the number of
different telephone lines through which the household could be reached,
and the number of eligible adults in the household. In Step 2, the initial
base weight was adjusted for differential nonresponse across the nine U.S.
Census divisions (New England, Middle Atlantic, East North Central, West
North Central, South Atlantic, East South Central, West South Central,
Mountain, and Pacific). In Step 3, the nonresponse-adjusted sampling
weights were poststratified to average population totals obtained from the
Current Population Survey (Bowler & Morisi, 2006) for the months during
which the present study was in the field (January 2002–June 2003).
Poststratification adjusts for known differences between the sample and
population on key variables that may be due to sampling error, undercov-
erage, or nonresponse. In Step 4, weight trimming was performed on the
final poststratified sampling weight for 5 interviewees whose value of their
poststratified weight exceeded four times the median value of the post-
stratified weight.
Respondent Characteristics
The respondent characteristics are described using weighted means and
percentages and the corresponding unweighted sample sizes. Fifty-three
percent (n ⫽ 1,277) of the participants were male and 47% (n ⫽ 1,529)
were female. Seventy-two percent (n ⫽ 2,190) were White, 13% (n ⫽ 328)
were Black, 8% (n ⫽ 159) were Hispanic, and 7% (n ⫽ 129) were of other
racial– ethnic makeup. The average age of participants was 39 years (N ⫽
2,806). In terms of highest level of education, 0.3% (n ⫽ 9) did not attend
high school; 4.0% (n ⫽ 110) attended high school but did not graduate;
23.5% (n ⫽ 637) graduated from high school or obtained a GED; 4.4%
(n ⫽ 116) attended trade, technical, or vocational training beyond high
school; 22.8% (n ⫽ 631) attended some college; 9.5% (n ⫽ 270) received
an associate’s degree; 19.6% (n ⫽ 570) received a bachelor’s degree; 3.2%
(n ⫽ 94) attended some graduate school; 10.0% (n ⫽ 292) received a
858
FRONE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

master’s degree; and 2.7% (n ⫽ 77) received a doctoral-level degree.
Average total family income was $63,224 (N ⫽ 2,806). Respondents had
an average of 1.82 (N ⫽ 2,806) financial dependents. Ten percent (n ⫽
291) of the respondents reported owning and operating their own business,
and 90% (n ⫽ 2,515) were wage and salary workers. In terms of the
Standard Occupational Classification (SOC; U.S. Office of Management
and Budget, 2000) major occupation groups, 7.9% (n ⫽ 293) were in
management occupations; 4.5% (n ⫽ 133) were in business and financial
operations management occupations; 3.0% (n ⫽ 90) were in computer and
mathematical science occupations; 2.4% (n ⫽ 61) were in architecture and
engineering occupations; 1.6% (n ⫽ 44) were in life, physical, and social
science occupations; 1.3% (n ⫽ 44) were in community and social services
occupations; 1.1% (n ⫽ 32) were in legal occupations; 7.4% (n ⫽ 219)
were in education, training, and library occupations; 2.5% (n ⫽ 71) were
in arts, design, entertainment, sports, and media occupations; 6.0% (n ⫽
177) were in health care practitioner and technical occupations; 3.5% (n ⫽
100) were in health care support occupations; 3.1% (n ⫽ 80) were in
protective services occupations; 4.2% (n ⫽ 117) were in food preparation
and serving occupations; 2.2% (n ⫽ 58) were in building and grounds
cleaning and maintenance occupations; 3.3% (n ⫽ 98) were in personal
care and service occupations; 9.0% (n ⫽ 254) were in sales and related
occupations; 15.8% (n ⫽ 459) were in office and administrative support
occupations; 0.5% (n ⫽ 13) were in farming, fishing, and forestry occu-
pations; 4.6% (n ⫽ 114) were in construction and extraction occupations;
4.6% (n ⫽ 111) were in installation, maintenance, and repair occupations;
5.3% (n ⫽ 146) were in production occupations; and 6.2% (n ⫽ 146) were
in transportation and material moving occupations. The participants
worked on average 42 hr (range ⫽ 2 hr– 60 or more hr, N ⫽ 2,806) per
week and held their present job for an average of 5 years (range ⫽ 1
month– 40 years, N ⫽ 2,806). In terms of work shifts, 76.8% (n ⫽ 2,173)
worked a fixed day shift, 7.4% (n ⫽ 195) worked a fixed evening shift,
3.4% (n ⫽ 83) worked a fixed night shift, 3.4% (n ⫽ 85) worked a rotating
shift, and 9.1% (n ⫽ 270) worked a nonstandard (irregular or flexible)
shift. Five and one half percent (n ⫽ 146) of the participants held seasonal
jobs, and 94.5% (n ⫽ 2,660) held nonseasonal jobs. Sixteen percent (n ⫽
429) belonged to a union, and 84% (n ⫽ 2,377) did not belong to a union.
General Demographic Characteristics
Gender was assessed with a single closed-ended question, and the
response options were coded 0 for women and 1 for men. Race– ethnicity
was assessed using two closed-ended questions. The first question asked
whether the respondents considered themselves to be Hispanic or Latino or
of Spanish origin, with response options of yes or no. The second question
asked the respondent about their race using the following six categories:
White, Black, Asian, American Indian or Alaska Native, Native Hawaiian
or other Pacific Islander, and mixed. For the present analyses, this infor-
mation was used to code respondents as follows: 0 ⫽ White (non-
Hispanic), 1 ⫽ Black (non-Hispanic), 2 ⫽ Hispanic, and 3 ⫽ other racial
groups. Age was coded in years from reported date of birth.
Years of formal education was assessed with a single closed-ended
question that used the following 10 response categories: 1 ⫽ less than high
school (Grades 1– 8); 2 ⫽ some high school without graduating; 3 ⫽ high
school graduate or GED; 4 ⫽ trade, technical, or vocational training
beyond high school; 5 ⫽ some college; 6 ⫽ associate’s degree; 7 ⫽
bachelor’s degree (e.g., BA, BS); 8 ⫽ some graduate school, 9 ⫽ master’s
degree (e.g., MA, MS, MBA); and 10 ⫽ doctoral-level degree (e.g., PhD,
MD, EdD, DBA, JD).
Number of financial dependents was assessed with an open-ended ques-
tion that asked about the number of people who currently rely on the
respondent financially regardless of whether they live with the respondent.
The responses were coded into six categories, ranging from 0 to 5 or more
financial dependents.
Occupational Demographic Characteristics
Type of worker was assessed with a single closed-ended question asking
whether each respondent operated his or her own business, professional
practice, or farm. A no response was coded 0 ⫽ wage and salary worker,
and a yes response was coded 1 ⫽ owner– operator.
Occupation was assessed with three open-ended questions that required
the respondents to (a) describe the kind of work they do, (b) provide their
formal job title, and (c) describe the duties or activities they do most often
on their job. All responses were entered verbatim by the interviewers into
the computer-assisted telephone interviewing program. These open-ended
responses were then coded into the 821 detailed 1998 SOC codes (U.S.
Office of Management and Budget, 2000). Each respondent’s occupational
information was coded independently by two coders who met weekly with
me to compare the two sets of codes completed during the previous week.
When a disagreement was found, all three individuals reexamined the
open-ended responses and the SOC descriptions and coding rules and
discussed the coding until a majority consensus was reached. For the
present report, the detailed SOC codes were first aggregated into the 22
SOC major occupation groups. At the level of the 22 SOC major occupa-
tion groups being used in this study, initial interrater agreement was 89.1%
for percentage agreement and 88.2% for Cohen’s kappa. However, because
of the relatively low base rate among the illicit drug use and impairment
variables and the fact that some of the 22 major occupation groups did not
contain a large number of individuals, I decided to further aggregate the
occupations. Although there are higher levels of aggregation for the SOC
occupation codes, they were not designed with studies of illicit drug use in
mind. In other words, these higher levels of aggregation may obscure the
relation between occupation and illicit drug use and impairment. Therefore,
past research on occupation and illicit drug use was consulted. The largest
and most detailed study was conducted by Hoffman et al. (1996) using
pooled data (N ⫽ 87,915) from multiple years (1991, 1992, 1993) of the
National Household Survey on Drug Abuse. This large sample allowed him
to estimate prevalence rates for more narrowly defined occupation groups.
The data from Hoffman et al.’s (1996) study showed that the majority of
occupations with elevated prevalence rates (defined as being at least 25%
higher than the sample mean) fell into the following five SOC major
occupation groups: (a) legal occupations; (b) arts, design, entertainment,
sports, and media occupations; (c) food preparation and serving related
occupations; (d) building and grounds cleaning and maintenance occupa-
tions; and (e) construction and extraction occupations. The remaining 17
major SOC occupations groups were combined into a single low-risk
occupation group. For the present regression analyses, five dummy vari-
ables were created representing each of the five high-risk occupations, with
the aggregate low-risk occupation group serving as the referent category.
Job tenure was assessed with a single open-ended item that required the
respondents to report the length of time in years and months they worked
in their current job, which was then converted into a continuous variable
representing number of years. Number of weekly work hours was assessed
with a single open-ended question that asked respondents to report the
number of hours per week they usually worked on their job, including paid
and unpaid overtime.
Work shift was coded into five categories on the basis of responses to
several questions. The first question was a closed-ended question that
required respondents to select the shift usually worked from several fixed
options. Each respondent also was asked a closed-ended question about
whether he or she usually began and ended his or her work shift about the
same time every day. Two open-ended questions required the respondent to
provide the time he or she usually started and ended his or her work shift.
Responses to these items were used to code respondents into the following
work shifts: (a) fixed day shift, (b) fixed evening shift, (c) fixed night shift,
(d) rotating shift, and (e) nonstandard (irregular or flexible) shift (e.g.,
Presser & Altman, 2002).
Seasonal job was assessed with a single closed-ended question that
asked whether the respondent worked in a seasonal job, which was defined
859
ILLICIT DRUG USE IN THE WORKFORCE AND WORKPLACE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

as working at certain times of the year because of weather conditions or
holidays. Responses were coded 0 for no and 1 for yes. Union membership
was assessed with a single closed-ended question that asked whether the
respondent belonged to a union. Responses were coded 0 for no and 1 for
yes.
Workforce Drug Use and Impairment
Two closed-ended questions were asked regarding the use of each of the
following six drugs: marijuana or hashish, cocaine or crack, sedatives,
tranquilizers, stimulants, and analgesics. Any use of marijuana or cocaine
represents illicit use. However, the use of the other four psychotherapeutic
drugs could be licit. Following standard procedures in prior national
surveys, interviewers asked respondents to report only nonmedical (i.e.,
illicit) use of sedatives, tranquilizers, stimulants, and analgesics. Nonmed-
ical use was defined as any use that is on your own (i.e., without a doctor’s
prescription, or in greater amounts than prescribed, or more often than
prescribed). For all six drugs, the first question assessed workforce use by
asking respondents how often they used each drug during the preceding 12
months. The second question assessed workforce impairment by asking
respondents how often they took enough of each drug to get high or stoned.
Measures of prevalence and frequency of use were computed for marijuana
use and impairment, cocaine use and impairment, any psychotherapeutic
drug use and impairment, and any illicit drug use and impairment. Col-
lapsing across the use of the four psychotherapeutic drugs is consistent
with typical official U.S. government statistics (Office of Applied Studies,
2003) on illicit drug use in the overall civilian population. The four
prevalence measures were scored 0 for no use–impairment and 1 for any
use–impairment. The four frequency measures were scored using the
6-point frequency response scale: 0 (never),1(less than monthly),2(1to
3 days per month),3(1 to 2 days per week),4(3 to 5 days per week), and
5(6 to 7 days per week).
Although other types of illicit drugs are used, such as inhalants, hallu-
cinogens, heroin, and methamphetamines, the present study focused on six
drugs that represented the three most commonly used categories of illicit
drugs (marijuana, cocaine, and psychotherapeutic drugs). Given the mul-
tiple goals of the National Survey of Workplace Health and Safety and the
detailed questions asked for each of these six drugs regarding use and
impairment in the workplace, it was not possible to ask about all potential
drugs of abuse. However, the omission of these other drugs should have
little effect on the overall prevalence rates. The reason is that the preva-
lence of using these other drugs is low, and most of the users of the omitted
drugs will use one or more of the drugs that were assessed. Thus, excluding
them should have little impact on the prevalence rates of any illicit drug
use. To examine the potential impact of not asking about these other four
categories of drugs on estimates of overall prevalence in the present study,
I conducted analyses using data from the 2002 National Survey on Drug
Use and Health (NSDUH; formerly called the National Household Survey
on Drug Abuse; Office of Applied Studies, 2003). In keeping with the
present study, the sample was restricted to currently employed (part and
full time) adults in the civilian workforce between the ages of 18 and 65
(N ⫽ 24,754). The data showed that of the 664 NSDUH respondents (2.7%
of the sample) who used inhalants, hallucinogens, heroin, or methamphet-
amines, 88.7% also used marijuana, cocaine, or psychotherapeutic drugs.
Therefore, excluding inhalants, hallucinogens, heroine, and methamphet-
amines had little impact on the overall prevalence estimate for any illicit
drug use in the workforce obtained from the 2002 NSDUH, reducing it
from 16.3% to 16.0%.
Workplace Drug Use and Impairment
Workplace drug use and impairment was assessed with 30 closed-ended
questions. To assess workplace drug use, I asked participants how often
during the past 12 months they used each of the six drugs described earlier
in each of the following four contexts: within 2 hr of starting their work
shift, during lunch breaks, during other breaks, and while working. To
assess workplace drug impairment, I asked participants how often they had
been at work high on or under the influence of each of the six drugs.
Measures of prevalence and frequency of use at work were computed for
marijuana use and impairment, cocaine use and impairment, any psycho-
therapeutic drug use and impairment, and any illicit drug use and impair-
ment. Also, prevalence measures were computed for using marijuana,
cocaine, any psychotherapeutic drugs, and any illicit drugs in each of the
four workplace contexts (within 2 hr of coming to work, during lunch
breaks, during other breaks, and while working). The prevalence measures
were scored 0 for no use–impairment and 1 for any use–impairment. The
frequency measures were scored using the 6-point frequency response
scale: 0 (never),1(less than monthly),2(1 to 3 days per month),3(1to
2 days per week),4(3 to 5 days per week), and 5 (6 to 7 days per week).
Data Analysis
Given the sampling design, the parameter estimates and inferential
statistics required the use of sampling weights (e.g., Lehtonen & Pahkinen,
2004). To obtain the various prevalence estimates and estimated population
totals, I computed weighted frequency distributions or weighted contin-
gency tables. To explore the relation of the four primary outcome variables
to the general and occupational demographic variables, I used weighted
bivariate and multiple ordered logistic regression analyses to obtain the
regression coefficients. In addition, the standard errors were based on
Taylor linearization, and overall model fit for the multiple ordered logistic
regression analyses was assessed with adjusted Wald F tests (e.g., Leh-
tonen & Pahkinen, 2004).
Results
Illicit Drug Use in the Workforce
Overall prevalence. The overall prevalence and frequency of
illicit drug use and impairment in the workforce is presented in
Table 1. The last column shows that during the prior 12 months,
11.33% of the workforce (14.2 million workers) used marijuana,
1.01% (1.3 million workers) used cocaine, 4.90% (6.2 million
workers) used psychotherapeutic drugs, and 14.06% (17.7 million
workers) used at least one illicit drug. In terms of using enough of
each type of drug to get high or stoned, 10.57% (13.3 million
workers) were impaired from marijuana, 0.93% (1.2 million work-
ers) were impaired from cocaine, 2.21% (2.8 million workers)
were impaired from psychotherapeutic drugs, and 11.23% (14.1
million workers) were impaired from any illicit drug. These overall
prevalence rates also show that marijuana is the illicit drug most
commonly used in the workforce.
Table 1 also presents data on the prevalence of workforce illicit
drug use and impairment by frequency of use. From the last two
rows for any illicit drug use or impairment, these data show that
using an illicit drug or being impaired by an illicit drug occurs
infrequently. This is made clearer by considering only those em-
ployees reporting some illicit drug use—68% did so three times
per month or less. This pattern for the frequency of any illicit drug
use or impairment was similar for marijuana and psychotherapeu-
tic drugs. However, cocaine showed a pattern of use and impair-
ment that was even more infrequent, with 90% of users doing so
three times per month or less.
Distribution. Table 2 shows the bivariate and multivariate
relations of frequency of illicit drug use and impairment in the
workforce to the general demographic and the occupational demo-
860
FRONE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

graphic characteristics. Beginning with the five general demo-
graphic characteristics, it can be seen that gender was related to
both use and impairment in the workforce. Men used illicit drugs
and were impaired by their use more often than women. Race was
unrelated to either outcome variable. Age and education were each
negatively related to both illicit drug use and impairment. Finally,
number of financial dependents was negatively related to illicit
drug use but was unrelated to impairment in the multivariate
analyses.
Turning to the results for the seven occupational characteristics,
the bivariate and multivariate results suggest that type of worker
was not related to workplace illicit drug use or impairment. Al-
though the bivariate analyses showed that job tenure, weekly work
hours, working the second shift, having a seasonal job, and union
membership were each related to both outcomes, the multivariate
analyses do not support these relations. Thus, the bivariate rela-
tions for these five occupational demographic characteristics are
likely to be spurious. The strongest and most consistent relations
involved occupation. Compared with the low-risk occupations,
illicit drug use and impairment in the workforce was higher among
the (a) arts, entertainment, sports, and media occupations and (b)
food preparation and serving occupations. With the exception of a
significant positive relation for overall illicit drug use in the
multivariate analyses, the legal occupations did not show consis-
tent evidence of being at risk for overall illicit drug use and
impairment. The bivariate relations indicated that the following
two occupations were at elevated risk for overall illicit drug use
and impairment: (a) building and grounds maintenance occupa-
tions and (b) construction and extraction occupations. However,
after I controlled for the other demographic characteristics, these
two occupational groups no longer showed elevated risk, suggest-
ing that their bivariate relations were spurious.
Illicit Drug Use in the Workplace
Overall prevalence. The overall prevalence and frequency of
illicit drug use and impairment in the workplace is presented in
Table 3. The last column shows that during the prior 12 months,
1.62% of the workforce (2 million workers) used marijuana, 0.13%
(169,000 workers) used cocaine, 1.80% (2.3 million workers) used
psychotherapeutic drugs, and 3.13% (3.9 million workers) used at
least one illicit drug at work. In terms of being at work under the
influence of an illicit drug, 1.74% (2.2 million workers) were
impaired from marijuana, 0.18% (233,000 workers) were impaired
from cocaine, 1.45% (1.8 million workers) were impaired from
psychotherapeutic drugs, and 2.88% (3.6 million workers) were
impaired from any illicit drug. These overall prevalence rates also
show that the use of marijuana and psychotherapeutic drugs are
equally prevalent in the workplace. Finally, comparing the prev-
alence of illicit drug use in the workplace (3.13%) with the
prevalence in the workforce (14.06%) reveals that 78% of workers
who use illicit drugs do not use them in the workplace.
Table 3 also presents data on the prevalence of workplace illicit
drug use and impairment by frequency of use. From the last two
rows for any illicit drug use or impairment, the data show that
among those employees who used an illicit drug at work, 44% used
3 days per month or less, and 59% were impaired at work 3 days
per month or less. This means that more than half (56%) of
employees who used an illicit drug at work and 41% of employees
who were impaired at work did so at least 1 day per week or more
often. Nonetheless, although a large percentage of employees who
used illicit drugs at work did so at least weekly, they represent only
1.8% of the total workforce.
Table 4 presents the workplace prevalence rates for each com-
bination of context of use and type of drug used. However, the last
column in Table 4 shows that 2.71% of the workforce (3.4 million
Table 1
Prevalence of Illicit Drug Use and Impairment in the Workforce During the Past 12 Months
Type of substance
Frequency of use
Never
Less than
monthly
1–3 days
per month
1–2 days
per week
3–5 days
per week
6–7 days
per week
Overall
prevalence
Marijuana use 88.67% 6.02% 1.98% 0.90% 0.93% 1.50% 11.33%
111,607,956 7,571,825 2,494,928 1,132,125 1,171,719 1,883,672 14,254,268
Marijuana impairment 89.43% 5.61% 1.77% 0.92% 1.06% 1.20% 10.57%
112,563,912 7,062,790 2,231,192 1,163,129 1,333,554 1,507,648 13,298,312
Cocaine use 98.99% 0.78% 0.14% 0.07% 0.02% 0.00% 1.01%
124,594,088 984,011 172,289 83,569 28,267 0 1,268,136
Cocaine impairment 99.07% 0.76% 0.08% 0.07% 0.02% 0.00% 0.93%
124,697,857 958,553 93,978 83,569 28,267 0 1,164,367
Psychotherapeutic drug use
a
95.10% 2.64% 0.87% 0.50% 0.59% 0.30% 4.90%
119,694,878 3,317,054 1,097,724 627,421 748,096 377,050 6,167,346
Psychotherapeutic drug impairment
a
97.80% 1.24% 0.41% 0.29% 0.22% 0.05% 2.21%
123,086,978 1,560,767 510,039 365,191 275,608 63,641 2,775,246
Any illicit drug use
b
85.94% 7.07% 2.50% 1.24% 1.49% 1.77% 14.06%
108,164,102 8,897,055 3,140,144 1,556,659 1,871,807 2,232,456 17,698,122
Any illicit drug impairment
b
88.77% 5.68% 1.95% 1.10% 1.26% 1.25% 11.23%
111,728,457 7,148,981 2,451,661 1,380,736 1,581,100 1,571,289 14,133,767
Note. N ⫽ 2,806. For each substance, the top number is the prevalence of use or impairment, and the bottom number is the estimated population total.
a
Psychotherapeutic drugs include sedatives, tranquilizers, stimulants, and analgesics.
b
Any illicit drugs include marijuana, cocaine, and the four
psychotherapeutic drugs.
861
ILLICIT DRUG USE IN THE WORKFORCE AND WORKPLACE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

workers) used illicit drugs within 2 hr of reporting to work, 1.82%
(2.3 million workers) used during lunch breaks, 1.19% (1.5 million
workers) used during other work breaks, and 1.72% (2.2 million
workers) used while working.
Distribution. Table 5 shows the bivariate and multivariate
relations of any illicit drug use and impairment in the workplace to
the general demographic and occupational demographic character-
istics. Beginning with the five general demographic characteris-
tics, it can be seen that gender was related to both use and
impairment in the workplace. Men used illicit drugs and were
impaired by their use at work more often than women. Regarding
race, individuals in the “other” category reported significantly
higher levels of workplace drug use and impairment than White
respondents and the other two racial groups (because the odds ratio
for other was larger than for Blacks and Hispanics) in the bivariate
analyses. However, the relation of the other racial group to work-
place impairment was not significant in the multivariate analyses.
On the whole, there is little evidence that workplace drug use and
impairment varies across the four racial groups. Age was nega-
tively related to both outcomes ( p ⬍ .07 for illicit drug impairment
in the multivariate results). Although education and number of
financial dependents were each negatively related to both out-
comes in the bivariate analyses, these four relations were not
significant in the multivariate analyses.
Turning to the results for the seven occupational characteristics,
the bivariate and multivariate results provide little evidence that
type of worker, work shift, holding a seasonal job, or union
membership were related to the two outcome variables. Although
the bivariate analyses showed that job tenure and weekly work
hours were each related to both outcomes, the multivariate analy-
ses do not support these relations. Thus, the bivariate relations for
these two occupational demographic characteristics are likely to be
spurious. The strongest and most consistent relations involved
occupation. Compared with the low-risk occupations, illicit drug
use and impairment in the workplace was higher among the (a)
legal occupations, (b) food preparation and serving occupations,
and (c) building and grounds maintenance occupations. The arts,
entertainment, sports, and media occupations were unrelated to
workplace illicit drug use and impairment in both the bivariate and
multivariate analyses. The bivariate analyses indicated that the
Table 2
Bivariate and Multiple Ordered Logistic Regression Results Predicting Illicit Drug Use and
Impairment in the Workforce From General and Occupational Demographic Characteristics
Predictor
Odds ratios
Illicit drug use Illicit drug impairment
Bivariate Multivariate Bivariate Multivariate
Gender (male) 1.42** 1.43* 1.54** 1.51**
Race
White RG RG
Black 1.14 1.06 1.11 1.05
Hispanic 1.16 0.85 1.26 0.95
Other 1.68 1.48 1.44 1.28
Age 0.94*** 0.95*** 0.94*** 0.95***
Education 0.88*** 0.91** 0.87*** 0.92*
Financial dependents 0.79*** 0.89* 0.79** 0.89
Type of worker (owner–operator) 0.92 1.13 0.89 0.98
Occupations at risk
Low risk RG RG
Legal 2.24 2.51* 1.04 1.29
Arts–entertainment–sports–media 3.92*** 3.43*** 4.56*** 4.33***
Food preparation and serving 4.54*** 2.78*** 4.13*** 2.39**
Building and grounds maintenance 2.36* 1.64 2.65* 1.87
Construction and extraction 2.21** 1.42 2.61*** 1.65
Job tenure 0.91*** 0.97 0.92**** 0.98
Weekly work hours 0.98*** 1.00 0.99* 1.00
Work shift
Days (1st) RG RG
Evenings (2nd) 1.62* 0.85 2.01* 1.18
Nights (3rd) 1.63 1.40 1.50 1.28
Rotating 0.98 0.62 1.05 0.73
Nonstandard (irregular–flexible) 1.37 1.04 1.39 1.07
Seasonal job (yes) 2.16*** 1.43 2.60*** 1.72*
Union member (yes) 0.48*** 0.70 0.49** 0.68
Overall model fit: Adjusted Wald F(21, 2785) 9.12*** 7.86***
Note. N ⫽ 2,806. An odds ratio greater than 1.0 indicates a positive relation between the predictor and ordered
outcome, an odds ratio of 1.0 indicates a null relation, and an odds ratio less than 1.0 indicates a negative relation.
For odds ratios greater or less than 1.0, greater deviations from 1.0 indicate increasing strength of the positive
and negative relations, respectively. RG ⫽ reference group.
* p ⱕ .05. ** p ⱕ .01. *** p ⱕ .001.
862
FRONE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

construction and extraction occupations were at elevated risk for
workplace illicit drug use and impairment. However, after control-
ling for the other demographic characteristics, the construction and
extraction occupations no longer showed significant elevated risk,
suggesting that these bivariate relations were spurious.
Subgroups at Highest Risk for Illicit Drug Use in the
Workforce and Workplace
The multivariate results presented earlier show that several
variables predicted the prevalence of illicit drug use and impair-
ment in the workforce or in the workplace. However, the three
most consistent predictors across all four outcomes were gender,
age, and working in certain occupations. To identify segments in
the employed population with the highest prevalence of illicit drug
use in the workforce and in the workplace, I crossed these three
demographic characteristics to create eight subgroups of workers.
Specifically, the eight subgroups were formed by crossing gender
(men vs. women), age (18 –30 vs. 31– 65), and occupation (low
risk for illicit drug use vs. high risk for illicit drug use). The cutoff
for age was based on a visual inspection of the bivariate relation
between age and the four outcome variables, which showed that
the prevalence of illicit drug use dropped off substantially after age
30. In this analysis, the high-risk occupation group contained four
of the five occupation groups presented in Tables 2 and 5. Because
the construction and extraction occupations were not related to any
of the four outcomes in the multivariate analyses, this occupation
category was combined with the other 17 occupation groups orig-
inally designated as low risk for illicit drug use. To do the sub-
group analyses, I report observed prevalence rates for each of the
Table 3
Prevalence of Illicit Drug Use and Impairment in the Workplace During the Past 12 Months
Type of substance
Frequency of use
Never
Less than
monthly
1–3 days
per month
1–2 days
per week
3–5 days
per week
6–7 days
per week
Overall
prevalence
Marijuana use at work 98.38% 0.27% 0.34% 0.47% 0.35% 0.18% 1.62%
123,829,432 344,387 433,635 590,744 435,032 228,994 2,032,792
Marijuana impairment at work 98.26% 0.63% 0.44% 0.13% 0.35% 0.18% 1.74%
123,673,596 792,825 556,724 169,268 442,990 226,822 2,188,628
Cocaine use at work 99.87% 0.07% 0.00% 0.06% 0.00% 0.00% 0.13%
125,693,173 85,482 0 83,569 0 0 169,051
Cocaine impairment at work 99.81% 0.80% 0.04% 0.00% 0.06% 0.00% 0.18%
125,628,867 95,842 53,946 0 83,569 0 233,357
Psychotherapeutic drug use at work
a
98.20% 0.77% 0.16% 0.44% 0.21% 0.22% 1.80%
123,591,410 967,509 202,541 549,381 271,599 279,783 2,270,814
Psychotherapeutic drug impairment at work
a
98.55% 0.56% 0.36% 0.18% 0.14% 0.20% 1.45%
124,038,159 709,967 451,275 229,667 181,661 251,495 1,824,065
Any illicit drug use at work
b
96.87% 0.86% 0.51% 0.82% 0.54% 0.40% 3.13%
121,927,509 1,079,050 636,176 1,032,142 678,570 508,777 3,934,715
Any illicit drug impairment at work
b
97.12% 1.04% 0.65% 0.27% 0.54% 0.38% 2.88%
122,232,988 1,311,765 817,312 341,682 680,159 478,317 3,629,235
Note. N ⫽ 2,806. For each substance, the top number is the prevalence of use or impairment, and the bottom number is the estimated population total.
a
Psychotherapeutic drugs include sedatives, tranquilizers, stimulants, and analgesics.
b
Any illicit drugs include marijuana, cocaine, and the four
psychotherapeutic drugs.
Table 4
Prevalence of Illicit Drug Use in the Workplace by Context of Use During the Past 12 Months
Context of use
Type of substance use
Marijuana Cocaine
Psychotherapeutic
drugs
a
Any illicit
drugs
b
Within 2 hr of coming to work 1.45% 0.13% 1.53% 2.71%
1,827,188 169,051 1,932,168 3,410,410
During lunch breaks 0.89% 0.05% 1.03% 1.82%
1,121,314 64,787 1,296,597 2,291,859
During other breaks 0.67% 0.07% 0.55% 1.19%
845,816 83,569 696,519 1,499,853
While working 0.71% 0.10% 1.03% 1.72%
896,622 120,295 1,295,771 2,161,361
Note. N ⫽ 2,806. For each substance, the top number is the prevalence of use or impairment, and the bottom
number is the estimated population total.
a
Psychotherapeutic drugs include sedatives, tranquilizers, stimulants, and analgesics.
b
Any illicit drugs include
marijuana, cocaine, and the four psychotherapeutic drugs.
863
ILLICIT DRUG USE IN THE WORKFORCE AND WORKPLACE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

four outcomes for each of the eight subgroups. Subgroup preva-
lence rates for illicit drug use and impairment in the workforce and
in the workplace were considered elevated if they were at least
50% (i.e., 1.5 times) greater than the rate in the overall sample and
are shown in boldface in Table 6. Subgroups 1– 4 in Table 6
represent women. It is clear that young women in high-risk occu-
pation groups (Group 2) have substantially elevated prevalence
rates on all four outcomes. Subgroups 5– 8 in Table 6 represent
men. Young men (Groups 5 and 6) have substantially elevated
prevalence on all four outcomes, but the prevalence rates are
especially high if these young men work in high-risk occupations
(Group 6).
Discussion
Illicit drug use (and impairment) in the workforce and in the
workplace are conceptually distinct and should not be used inter-
changeably in empirical research or in policy discussions. How-
ever, as discussed earlier, not only does little detailed research
exist on the prevalence, frequency, and distribution of illicit drug
use in the workforce but even less research exists on the preva-
lence, frequency, and distribution of illicit drug use in the work-
place. This distinction is important for the design of research
looking at the putative causes and outcomes of employee drug use.
Moreover, research on both dimensions of illicit drug use and
impairment is important for managers and policymakers. Thus, an
important goal of this national study was to begin exploring in
detail the prevalence, frequency, and distribution of illicit drug use
in the workforce and in the workplace, with an eye toward iden-
tifying those segments of the workforce at greatest risk. Another
important goal was to discuss the specific implications of these
findings for future theoretically based research and for manage-
ment practice and policy. But before turning to implications, I will
discuss potential methodological limitations of this study.
Study Limitations
Two potential study limitations should be noted. The first lim-
itation is the potential for nonresponse bias when response rates
fall short of 100%. However, unit nonresponse (i.e., a response rate
less than 100%) is a necessary but not sufficient condition for
nonresponse bias. Nonresponse bias also requires that the reason
Table 5
Bivariate and Multiple Ordinal Logistic Regression Results Predicting Illicit Drug Use and
Impairment in the Workplace From General and Occupational Demographic Characteristics
Predictor
Odds ratios
Illicit drug use Illicit drug impairment
Bivariate Multivariate Bivariate Multivariate
Gender (male) 2.46** 2.45** 2.82** 2.98**
Race
White RG RG RG RG
Black 0.89 0.89 0.59 0.55
Hispanic 0.89 0.62 0.84 0.57
Other 3.71** 3.01* 3.13* 2.30
Age 0.93*** 0.97* 0.93*** 0.97
Education 0.86* 0.91 0.87* 0.91
Financial dependents 0.67** 0.80 0.71** 0.85
Type of worker (owner–operator) 0.97 1.87 1.23 2.30*
Occupations at risk
Low risk RG RG RG RG
Legal 6.11* 3.90* 6.84* 4.71*
Arts–entertainment–sports–media 2.57 1.90 2.88 1.90
Food preparation and serving 9.43*** 6.29*** 9.56*** 6.37***
Building and grounds maintenance 5.52** 3.45* 5.80*** 3.29*
Construction and extraction 3.91*** 2.13 3.97*** 2.09
Job tenure 0.85* 0.93 0.86* 0.93
Weekly work hours 0.97** 0.99 0.97** 0.98
Work shift
Days (1st) RG RG RG RG
Evenings (2nd) 1.90 0.82 2.08 0.91
Nights (3rd) 2.67 2.06 3.19 2.74
Rotating 0.46 0.28 0.80 0.60
Nonstandard (irregular–flexible) 0.92 0.64 1.22 0.87
Seasonal job (yes) 2.05 1.21 2.39* 1.29
Union member (yes) 0.44 0.63 0.49 0.71
Overall model fit: Adjusted Wald F(21, 2785) 5.01*** 5.12***
Note. N ⫽ 2,806. An odds ratio greater than 1.0 indicates a positive relation between the predictor and ordered
outcome, an odds ratio of 1.0 indicates a null relation, and an odds ratio less than 1.0 indicates a negative relation.
For odds ratios greater or less than 1.0, greater deviations from 1.0 indicate increasing strength of the positive
and negative relations, respectively. RG ⫽ reference group.
* p ⱕ .05. ** p ⱕ .01. *** p ⱕ .001.
864
FRONE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

for nonparticipation is associated with the substantive variables of
interest in any study or report (e.g., Groves et al., 2004). This
further means that the potential for nonresponse bias may not be
similar across variables in the same study. So each set of variables
used from the same data set needs to be evaluated separately.
Although one can never rule out potential nonresponse bias with
absolute certainty, there is little reason to expect that nonresponse
bias had a major effect on the prevalence rates reported in this
study. Two types of unit nonresponse need to be considered. First,
there are potentially eligible households that could not be con-
tacted even with the 20 attempts used in this study. However,
Groves et al. (2004) suggested that most noncontact nonresponse
is unlikely to be related to the purpose of a study. Second, there are
eligible households that refuse to be screened for eligibility, and
there are eligible individuals who refuse to participate even with
efforts at refusal conversion. However, refusal nonresponse is
unlikely to be associated with the drug use variables in this study.
All households refusing to be screened and most eligible individ-
uals who refused to participate did so before the informed consent
statement could be read to them. It was only during informed
consent that eligible individuals were given a general description
of the various types of issues that would be covered. Moreover,
Table 6
Prevalence of Illicit Drug Use and Impairment in the Workforce and Workplace for
Combinations of Leading Risk Factors
Leading risk
factor
Illicit drug use
in the workforce
Illicit drug
impairment
in the workforce
Illicit drug use
in the workplace
Illicit drug
impairment in
the workplace
Overall sample 14.1% 11.2% 3.1% 2.9%
17,698,122 14,133,767 3,934,715 3,629,235
Group 1 19.5% 13.1% 2.1% 0.9%
2,519,655 1,696,512 267,063 119,308
Gender: women
Age: 18–30
OCC: low risk
Group 2 43.4% 42.6% 10.6% 11.4%
1,119,814 1,099,165 273,116 293,453
Gender: women
Age: 18–30
OCC: high risk
Group 3 6.8% 5.0% 0.8% 0.8%
2,695,413 1,976,745 333,931 317,761
Gender: women
Age: 31–65
OCC: low risk
Group 4 16.6% 14.1% 4.6% 3.7%
666,012 562,772 182,173 149,866
Gender: women
Age: 31–65
OCC: high risk
Group 5 24.7% 20.9% 7.7% 6.8%
4,055,756 3,423,538 1,270,736 1,110,253
Gender: men
Age: 18–30
OCC: low risk
Group 6 55.8% 37.2% 28.0% 26.3%
1,757,038 1,173,121 882,262 828,546
Gender: men
Age: 18–30
OCC: high risk
Group 7 10.2% 8.8% 1.5% 1.7%
4,553,036 3,894,399 656,932 741,546
Gender: men
Age: 31–65
OCC: low risk
Group 8 11.7% 10.8% 2.4% 2.4%
331,398 307,515 68,502 68,502
Gender: men
Age: 31–65
OCC: high risk
Note. N ⫽ 2,806. For each group, the top number is the prevalence of use or impairment, and the bottom
number is the estimated population total. Boldface indicates that subgroup prevalence rates for illicit drug use
and impairment in the workforce and in the workplace were elevated if they were at least 50% (i.e., 1.5 times)
greater than the rate in the overall sample. OCC ⫽ occupations at low and high risk for illicit drug use (see
Method section).
865
ILLICIT DRUG USE IN THE WORKFORCE AND WORKPLACE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

when households were contacted to ascertain whether an eligible
individual resided in the household, the working title of the study
was the National Survey of Workplace Health and Safety, which
does not imply an emphasis on illicit drug use. The expectation
that refusal nonresponse was unlikely to have had an impact on the
prevalence estimates of illicit drug use from the present study is
supported by a national study of illicit drug use in the workforce of
registered nurses (N ⫽ 4,438) conducted by Trinkoff and Storr
(1997). These researchers used a mail survey whereby the original
mailing was followed by four additional mailings, the last being
sent by certified mail. The response rate increased from 43% after
the initial mailing to 78% after the final mailing. However, their
results revealed that the prevalence rates for illicit drug use did not
differ significantly across the five waves of mailings. Moreover,
the workforce prevalence rates based on the final sample (3.5% for
marijuana– cocaine and 6.6% for psychotherapeutic drugs) were
similar to the prevalence rates obtained after the first mailing
(3.1% for marijuana– cocaine and 6.4% for psychotherapeutic
drugs).
The second potential limitation in this study was the use of
self-reports of illicit drug use and impairment. Although it is naive
to assume all self-reports are veridical, Turkkan (2000) and Bald-
win (2000) pointed out that with behaviors that can be hidden, such
as drug use, there may be no better measurement methods. Turk-
kan further noted that “biologic and other data can be as prone to
false positives, false negatives, and other inaccuracies as self-
report data” (p. 1). One alternate avenue to obtain information
about a person’s use of illicit drugs is the use of collateral reports
of people who know the respondent. However, substance use can
be hidden from other observers, and collaterals are subject to the
same types of biases as the target respondent. In fact, research
shows that self-reports of substance use and other illicit behaviors
tend to be reliable and valid, and there is no reason to believe that
reports from collateral individuals will be more reliable or more
valid (e.g., Connors & Maisto, 2003; Darke, 1998). Another ave-
nue is to assess the use of illicit drugs through biologic tests (i.e.,
urine tests). However, biologic tests can only determine drug use
that has occurred over a short (a few hours to a few months) period
of time. If one is interested in rates of illicit drug over some
extended period, such as 3–12 months, the short window of bio-
logic tests may lead to underestimated prevalence rates because
they will misclassify infrequent users. Cook and Bernstein (1994)
compared the 6-month prevalence of workforce illicit drug use by
self-report to prevalence estimates obtained by urinalysis in a
sample of 621 employed adults. The overall agreement rate be-
tween these two methods was 89.4%, with self-reports providing a
slightly higher prevalence rate (10.3%) than urinalysis (8.0%).
Another limitation of biologic tests is that they cannot provide
information on the frequency of using illicit drugs. A final limi-
tation that is central to the present study is that biologic tests
cannot determine the context in which the drugs were used. Thus,
they cannot be used to explore the prevalence and frequency of
drug use in the workplace.
Finally, as an additional check for potential nonresponse bias
and potential self-report bias, the overall prevalence of illicit drug
use in the workforce from this study was compared with the
estimate obtained from the 2002 NSDUH. The NSDUH is a large
national household survey conducted annually to track illicit drug
use in the U.S. population. The 2002 NSDUH had a high response
rate (79%) and used audio computer-assisted self-interviewing
technology on laptop computers to minimize the underreporting of
substance use (e.g., Tourangeau, Rips, & Rasinski, 2000). Using
the same sample selection criteria used in the present study yielded
a subsample of 24,754 respondents from the 2002 NSDUH. The
overall prevalence of illicit drug use (i.e., marijuana, cocaine, and
any psychotherapeutic drugs) in the workforce from the present
study (14.1%) was close to that found in the 2002 NSDUH
(16.0%).
Implications and Directions for Future Research and
Theory
The results from this study show that the prevalence of illicit
drug use in the workforce and in the workplace are sufficiently
high, especially in some subgroups, that researchers, managers,
and policymakers should be interested in their causes and out-
comes. Unfortunately, the workplace causes and outcomes of illicit
drug use among employees are still not well understood (for
reviews, see Frone, 2004; Harris, 2004). This is partly due to lack
of research attention and partly because of important limitations in
the research conducted to date. In the remainder of this section, I
will highlight the implications of the present results for future
theoretically driven research by organizational researchers.
Sampling. Whether one assesses the prevalence of illicit drug
use, the frequency of use, or amount of a drug consumed, the
present data suggest that research using general samples of work-
ing individuals may not be an efficient strategy because of strong
floor effects. For illicit drug use in the workforce, 86% of the
sample will have a score of zero. For illicit drug use in the
workplace, 97% of the sample will have a score of zero. Therefore,
in general samples, the drug use variables will be highly skewed
and have little overall variation. Said differently, in a general
sample, substantial resources will be devoted to collecting data
from a preponderance of respondents who will have variation on
the putative causes and outcomes of illicit drug use but no varia-
tion on illicit drug use. This leads to low power to detect the causes
and outcomes of employee substance use, unless the general sam-
ples are much larger than typically seen in organizational studies.
Failure to detect causal relations because of low power undermines
the validation and development of theoretical explanations for the
causes and outcomes of employee substance use. The results in
Table 6 provide the first detailed data for researchers in terms of
sample selection. Clearly, it would be a better use of scarce
resources to begin a study of the predictors or outcomes of illicit
drug use in the workforce and in the workplace by sampling within
the segments of the workforce at greatest risk of illicit drug use.
The advantages are twofold. First, there would be more variation
in the illicit drug use variables. Moreover, even in the subgroup
with the highest rates of illicit drug use (i.e., drug use in the
workforce among young men in at-risk occupations), there is a
sufficient percentage of nonusers to minimize any restriction of
range problems on the various workplace causes and outcomes of
drug use. Second, research would be conducted on the predictors
and outcomes of illicit drug use in those segments of the workforce
in which such knowledge can do the most good in terms of
developing prevention programs to minimize illicit drug use and
its potential impact on employees and employers.
866
FRONE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Causes. Past research on the causes of employee illicit drug
use has been fairly limited. The predictors of illicit drug use in the
workforce have received some attention, though very little re-
search has explored the predictors of illicit drug use in the work-
place (see Frone, 2003, for a recent exception). An immediate need
for future research is to look at a broader set of predictors and
theoretical models regarding the putative causes of illicit drug use
among employees and to differentiate between illicit drug use in
the workforce and in the workplace (Frone, 1999).
The present data showed that certain occupations had higher
prevalence rates for illicit drug use in the workforce and in the
workplace than other occupations. This finding lends indirect
evidence that work environments might represent an underlying
cause of employee illicit drug use. The conceptual literature on
employee substance use suggests that three general paradigms can
be used to organize the major workplace predictors of illicit drug
use (Frone, 1999; Harris, 2004). The first is the social control
paradigm, which comes out of the general deviance literature (e.g.,
Shoemaker, 1996). Applied to work organizations, this paradigm
proposes that employees who are not integrated into or regulated
by the work organization are at higher risk of using illicit drugs.
The specific work conditions that put employees at risk include
low levels of commitment or attachment to an organization, high
mobility during work hours, low visibility of work behaviors,
working in isolation, low levels of supervision, and a lack of
formal and informal polices and disciplinary actions regarding
illicit drug use. The present findings show that building and
grounds maintenance occupations are at elevated risk for illicit
drug use at work. This occupation group may also be more likely
to have low visibility of work behaviors and to work in isolation.
Some research has recently begun to explore a few of the organi-
zational conditions implicated by social control theory in relation
to illicit drug use among employees (e.g., Frone, 2003; Mac-
Donald, Wells, & Wild, 1999). However, the measures used have
been narrow and do not provide an adequate test of the relation of
social control to employee drug use.
The second general paradigm is the availability paradigm, which
suggests that work settings in which illicit drugs are physically or
socially available may promote illicit drug use among employees.
Building from general availability theories of alcohol use, Ames
and colleagues (Ames & Janes, 1992; Ames & Grube, 1999)
identified and defined two dimensions of overall alcohol availabil-
ity and workplace alcohol availability. Although developed in the
area of alcohol studies, these conceptions of availability apply to
illicit drug use as well. The first dimension is physical availability
of illicit drugs, which refers to the ease of obtaining illicit drugs
either overall or in the workplace (Ames & Grube, 1999). The
second dimension is social availability of illicit drugs either overall
or in the workplace. Specifically, social availability refers to
general normative support for illicit drug use and has two compo-
nents (Ames & Grube, 1999). The first component represents the
use of illicit drugs by members of one’s overall or workplace
social network (i.e., descriptive norms). The second component
represents normative approval or disapproval of illicit drug use by
members of one’s overall or workplace social network (i.e., in-
junctive norms). The present findings might suggest that all of the
high-risk occupations, compared with the low-risk occupations,
have higher levels of both physical and social availability of illicit
drugs. Yet little empirical research to date has explored potential
causes of employee illicit drug use derived from theories of phys-
ical and social availability. Recent exceptions include studies by
Frone (2003), MacDonald et al. (1999), and Lehman, Farabee, and
Bennett (1998). However, the measures used in past research have
been narrow and do not provide an adequate test of the relation of
either physical or social availability to employee drug use, either in
the workforce or in the workplace.
The third paradigm is the alienation–stress paradigm, which
states that illicit drug use among employees may be a response to
adverse physical and psychosocial qualities of the work environ-
ment. In other words, illicit drug use may be used in an effort to
regulate negative emotions or thoughts that results from adverse
work environments. A number of studies have explored the
alienation–stress paradigm in terms of employee alcohol use (see
Cooper, Russell, & Frone, 1990; Frone, 1999, for reviews). How-
ever, much less research has focused on illicit drug use among
employees. Frone (1999) discussed a number of conceptual models
(i.e., simple cause– effect model, mediation model, moderation
model, and moderated–mediation model) that can be used to frame
future research and provided several suggestions for the design of
future research on the alienation–stress paradigm. For example,
given the many potential sources of work stress (e.g., Barling,
Kelloway, & Frone, 2005), attention needs to be devoted to dis-
covering which stressors may be the strongest and most consistent
causes of employee substance use and impairment in the work-
force and in the workplace. In terms of building and testing more
detailed mediational models of work stress and illicit drug use in
the workforce and the workplace, researchers can draw on affect
regulation theory (e.g., Cooper, Frone, Russell, & Mudar, 1995).
Finally, moderators of the relation between work stress and illicit
drug use need to receive more attention.
Outcomes. The mere fact that workers use illicit drugs or
achieve some level of impairment off and on the job is not
sufficient evidence that employee productivity will be affected
adversely. However, the prevalence rates observed in this study,
especially in certain subgroups, are a sufficient reason to explore
the relation between illicit drug use and productivity in more
detail. Unfortunately, past research on illicit drug use and produc-
tivity is not only in short supply, it suffers from numerous con-
ceptual and methodological problems (Frone, 2004; Harris, 2004).
Also, the focus of past research is somewhat uneven. Most re-
search has focused on absenteeism, some research has focused on
injuries and accidents, and little research has focused on issues like
task performance, contextual performance, interpersonal problems
at work, fatigue and sleeping on the job, and aggressive behavior
at work.
Because of conceptual and methodological limitations, past
empirical research shows little evidence of consistent and robust
relations between illicit drug use and employee productivity (for
reviews, see Frone, 2004; Normand et al., 1994). In an attempt to
put the weak and inconsistent findings of past research into per-
spective, Frone (2004) developed a general conceptual model of
employee substance use and productivity. In this model, a number
of issues are represented that have been ignored in past research.
The most important limitation in past field research is the failure to
match the context of illicit drug use to the type of productivity
outcome. In terms of context, illicit drug use can occur off the job
or on the job. Productivity outcomes are of two general types—
attendance outcomes (absenteeism and tardiness) and performance
867
ILLICIT DRUG USE IN THE WORKFORCE AND WORKPLACE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

outcomes (e.g., task performance, contextual performance, coun-
terproductive behavior, workplace injuries and accidents). Past
research has, with few exceptions, only assessed illicit drug use in
the workforce, which primarily represents off-the-job drug use.
However, only illicit drug use and impairment in the workplace
(i.e., on-the-job use) can be expected to have a major impact on
behaviors and outcomes that occur at work (Frone, 1998). In
contrast, illicit drug use and impairment in the workforce, or
off-the-job use and impairment, is likely to have its major effect on
attendance behaviors. The second issue that has not received
sufficient attention is the distinction between drug use and impair-
ment. Impairment (i.e., intoxication), which can occur off the job
and on the job, is likely to be the most proximal predictor of
employee productivity outcomes and mediates between illicit drug
use and the outcomes. In addition, the relations between illicit drug
use and impairment and between impairment and productivity
outcomes are likely to be moderated by a number of psychological,
behavioral, and physiological processes (see Frone, 2004, for a
detailed discussion). However, past research has not explored
potential moderating processes.
Another interesting avenue for future research is to examine the
morale and productivity of employees who do not use drugs at
work. Most research has been concerned with the work outcomes
of the person using drugs. However, because this study shows that
workplace substance use may be prevalent in certain segments of
the workforce, it would be useful to see whether exposure to
individuals who use drugs at work or arrive at work impaired has
a negative impact on the morale and performance of coworkers
who do not use drugs at work or come to work impaired. If it does,
the impact of workplace drug use on productivity may be broader
than what is typically assumed.
Implications for Management Practice and Policy
An important issue regarding management practice and policy
surrounding illicit drug use is the growing use of various types of
drug testing (e.g., preemployment, random, for cause) in U.S.
organizations (Fendrich & Kim, 2002; Hartwell, Steele, French, &
Rodman, 1996). Testing employees for illicit drug use has pre-
sumably become more commonplace in U.S. organizations as a
means of minimizing on-the-job safety and performance problems
that may result from the use of illicit drugs. Therefore, it is
important for researchers and managers to evaluate the utility of
drug testing. In order to evaluate fully the value of drug testing as
a general management technology, three questions need to be
addressed. First, is the prevalence of illicit drug use and impair-
ment in the workplace high enough to warrant testing? Second, is
illicit drug use and impairment in the workplace related to on-the-
job safety and performance outcomes? Third, can drug testing
identify individuals who are likely to use illicit drugs at work or
who will be at work under the influence of an illicit drug? The
present study can address the first question. The second and third
questions cannot be directly addressed with the present data, and
they have not been addressed adequately in past research. Regard-
ing the prevalence issue, this study found that the overall preva-
lence of illicit drug use and impairment in the workplace was
3.13% and 2.88%, respectively. This represents approximately
3.6 –3.9 million workers in a total workforce of about 126 million
workers. Thus, for most employers, illicit drug use in the work-
place should not be a major concern. Nonetheless, as shown in
Table 6, there are three segments of the workforce (Groups 2, 5,
and 6) in which the prevalence of workplace illicit drug use and
impairment is substantially higher, ranging from 6.8% to 28.0%.
These segments of the workforce should be of special concern for
employers whose employees heavily comprise such individuals.
For such employers, the prevalence of workplace illicit drug use is
high enough to warrant drug testing programs if future research
addressing the other two questions noted above can show that (a)
workplace drug use is related to poor safety and performance and
(b) drug testing can identify individuals who will use drugs at work
or arrive at work impaired. Nonetheless, in terms of general policy
on drug testing for the U.S. workforce, the elevated prevalence
rates in the three at-risk subgroups need to be placed into context.
The estimated population of all individuals falling into these three
subgroups represents 17.6% of the U.S. workforce, and the esti-
mated population of individuals in these three subgroups reporting
workplace illicit drug use represents 1.9% of the U.S. workforce.
Conclusion
The subtitle of Newcomb’s (1994) article on the prevalence of
workplace substance use was “Cause for Concern or Irrational
Hysteria?” Although the present data suggest that there is some
cause for concern in specific segments of the U.S. workforce, there
is little justification for widespread hysteria. Furthermore, the
ability of managers to develop defensible and effective interven-
tions is contingent on knowledge of the prevalence and distribution
of illicit drug use in the workforce and in the workplace and on
better theoretically based research on the putative causes and
productivity outcomes of illicit drug use among employees. It is
hoped that this study will heighten the motivation of industrial–
organizational psychologists and other organizational researchers
to begin developing a new generation of integrative theoretical
research on the causes and outcomes of illicit drug use in the
workforce and in the workplace.
References
American Psychiatric Association. (1994). Diagnostic and statistical man-
ual of mental disorders (4th ed.). Washington, DC: Author.
Ames, G., & Janes, C. J. (1992). A cultural approach to conceptualizing
alcohol and the workplace. Alcohol Health and Research World, 16,
112–119.
Ames, G. M., & Grube, J. W. (1999). Alcohol availability and workplace
drinking: Mixed method analyses. Journal of Studies on Alcohol, 60,
383–393.
Ames, G. M., Grube, J. W., & Moore, R. S. (1997). The relationship of
drinking and hangovers to workplace problems: An empirical study.
Journal of Studies on Alcohol, 58, 37–47.
Baldwin, W. (2000). Information no one else knows: The value of self-
report. In A. A. Stone, J. S. Turkkan, C. A. Bachrach, J. B. Jobe, H. S.
Kurtzman, & V. S. Cain (Eds.), The science of self-report: Implications
for research and practice (pp. 3–7). Mahwah, NJ: Erlbaum.
Barling, J., Kelloway, E. K., & Frone, M. R. (Eds.). (2005). Handbook of
work stress. Thousand Oaks, CA: Sage.
Bowler, M., & Morisi, T. L. (2006). Understanding the employment
measures from the CPS and CES survey. Monthly Labor Review, 129,
23–38.
Connors, G. C., & Maisto, S. A. (2003). Drinking reports from collateral
individuals. Addiction, 98(Suppl. 2), 21–29.
868
FRONE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Cook, R. F., & Bernstein, A. (1994). Assessing drug use prevalence in the
workplace: A comparison of self-report and urinalysis. International
Journal of the Addictions, 29, 1057–1068.
Cooper, M. L., Frone, M. R., Russell, M., & Mudar, P. (1995). Drinking to
regulate positive and negative emotions: A motivational model of alco-
hol use. Journal of Personality and Social Psychology, 69, 990 –1005.
Cooper, M. L., Russell, M., & Frone, M. R. (1990). Work stress and
alcohol effects: A test of stress-induced drinking. Journal of Health and
Social Behavior, 31, 260 –276.
Darke, S. (1998). Self-report among injection drug users: A review. Drug
and Alcohol Review, 51, 253–263.
Drug-Free Workplace Act of 1988, Pub. L. No. 100 –690, 102 Stat. 4304.
Fendrich, M., & Kim, J. Y. S. (2002). The experience and acceptability of
drug testing: Poll trends. Journal of Drug Issues, 32, 81–96.
Frone, M. R. (1998). Predictors of work injuries among employed adoles-
cents. Journal of Applied Psychology, 83, 565–576.
Frone, M. R. (1999). Work stress and alcohol use. Alcohol Research and
Health, 23, 284 –291.
Frone, M. R. (2003). Predictors of overall and on-the-job substance use
among young workers. Journal of Occupational Health Psychology, 8,
39 –54.
Frone, M. R. (2004). Alcohol, drugs, and workplace safety outcomes: A
view from a general model of employee substance use and productivity.
In J. Barling & M. R. Frone (Eds.), The psychology of workplace safety
(pp. 127–156). Washington, DC: American Psychological Association.
Gleason, P. M., Veum, J. R., & Pergamit, M. R. (1991). Drug and alcohol
use at work: A survey of young workers. Monthly Labor Review, 8, 3–7.
Groves, R. M., Fowler, F. J., Couper, M. P., Lepkowski, J. M., Singer, E.,
& Tourangeau, R. (2004). Survey methodology. New York: Wiley.
Harris, M. M. (2004). Alcohol and drug use in the workplace. In R. W.
Griffin & A. O’Leary-Kelly (Eds.), The dark side of organizational
behavior (pp. 341–372). New York: Pfeiffer.
Hartwell, T. D., Steele, P. D., French, M. T., & Rodman, N. F. (1996).
Prevalence of drug testing in the workplace. Monthly Labor Review, 119,
35– 42.
Hoffman, J. P., Brittingham, A., & Larison, C. (1996). Drug use among
U.S. workers: Prevalence and trends by occupation and industry cate-
gories (DHHS Pub. No. SMA 96 –3089). Washington, DC: U.S. Gov-
ernment Printing Office.
Hoffman, J. P., Larison, C., & Sanderson, A. (1997). An analysis of worker
drug use and workplace policies and programs (DHHS Pub. No. SMA
97–3142). Washington, DC: U.S. Government Printing Office.
Korn, E., & Graubard, B. (1999). Analysis of health surveys. New York:
Wiley.
Lehman, W. E. K., Farabee, D. J., & Bennett, J. B. (1998). Perceptions and
correlates of co-worker substance use. Employee Assistance Quarterly,
13, 1–22.
Lehtonen, R., & Pahkinen, E. (2004). Practical methods for design and
analysis of complex surveys (2nd ed.). New York: Wiley.
Levy, P. S., & Lemeshow, S. (1999). Sampling of populations: Methods
and applications (3rd ed.). New York: Wiley.
MacDonald, S., Wells, S., & Wild, T. C. (1999). Occupational risk factors
associated with alcohol and drug problems. American Journal of Drug
and Alcohol Abuse, 25, 351–369.
Newcomb, M. D. (1994). Prevalence of alcohol and other drug use on the
job: Cause for concern or irrational hysteria? Journal of Drug Issues, 24,
403– 416.
Normand, J., Lempert, R. O., & O’Brien, C. P. (1994). Under the influ-
ence? Drugs and the American work force. Washington, DC: National
Academy Press.
Office of Applied Studies. (2003). Results from the 2002 National Survey
on Drug Use and Health: National findings (DHHS Pub. No. SMA
03–3836). Rockville, MD: Substance Abuse and Mental Health Services
Administration.
Potthoff, R. F. (1994). Telephone sampling in epidemiologic research: To
reap the benefits, avoid the pitfalls. American Journal of Epidemiology,
139, 967–978.
Presser, H. B., & Altman, B. (2002). Work shifts and disability: A national
view. Monthly Labor Review, 125, 11–24.
Roman, P. M., & Blum, T. C. (1995). Employers. In R. H. Coombs &
D. M. Ziedonis (Eds.), Handbook on drug abuse prevention (pp. 139 –
158). Boston: Allyn & Bacon.
Schaufeli, W. B. (2004). The future of occupational health psychology.
Applied Psychology: An International Review, 53, 502–517.
Shoemaker, D. J. (1996). Theories of delinquency (3rd ed.). New York:
Oxford University Press.
Sindelar, J. (1998). Social costs of alcohol. Journal of Drug Issues, 28,
763–780.
Tourangeau, R., Rips, L. J., & Rasinski, K. (2000). The psychology of
survey response. New York: Cambridge University Press.
Trinkoff, A. M., & Storr, C. L. (1997). Collecting substance use data with
an anonymous mailed questionnaire. Drug and Alcohol Dependence, 48,
1– 8.
Tucker, C., Lepkowski, J. M., & Piekarski, L. (2002). The current effi-
ciency of list-assisted telephone sampling designs. Public Opinion
Quarterly, 66, 321–338.
Turkkan, J. S. (2000). General issues in self-report. In A. A. Stone, J. S.
Turkkan, C. A. Bachrach, J. B. Jobe, H. S. Kurtzman, & V. S. Cain
(Eds.), The science of self-report: Implications for research and practice
(pp. 1–2). Mahwah, NJ: Erlbaum.
U.S. Office of Management and Budget. (2000). Standard occupational
classification manual. Lanham, MD: Bernan Press.
Wittchen, H.-U. (2004). Continued needs for epidemiological studies of
mental disorders in the community. Psychotherapy and Psychosomatics,
73, 197–206.
Received February 1, 2005
Revision received July 11, 2005
Accepted July 13, 2005 䡲
869
ILLICIT DRUG USE IN THE WORKFORCE AND WORKPLACE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.