AMERICAN JOURNAL OF INDUSTRIAL MEDICINE 45:297–304 (2004)

Overcoming Recruitment Challenges in

Construction Safety Intervention Research

1
Pamela Kidd, PhD, FNP-C, FAAN, Mark Parshall, PhD, RN,2
Susan Wojcik, MS,
3

and Tim Struttmann, MsPH4

Background Recruiting workers in small construction companies and securing their par-
ticipation in voluntary safety programs or safety research poses unique challenges. Worker
turnover and worksite changes contribute to difficulties in locating and enrolling parti-
cipants. Economic pressures and time demands potentially threaten ongoing participation.
Methods Six simulation exercises designed to reduce back and fall injuries in small
construction companies were developed based on data from focus groups of workers and
company owners. Working with a workers’ compensation insurer, we had access to owner-
operators of general, heavy, and special trade construction companies reporting less than
$10,000 in payroll expenses. Recruitment methods included a participation incentive,
mailed invitations followed by phone contacts, and follow-up reminders.
Results Despite using recruitment methods recommended in the literature, participation
rates were low over a 2-year intervention period. Because of these difficulties, factors
affecting participation or nonparticipation became an additional research focus. Owners’
perceptions of already having a good safety record and of the time demands of parti-
cipation were the most commonly cited reasons for not participating.
Conclusions Literature on recruitment emphasizes processes and procedures under
investigator control rather than understanding potential participants’ judgments about the
adequacy of their existing practices and the potential benefits of intervention participation
relative to potential time and productivity trade-offs. Greater attention to such judgments
may enhance recruitment and participation in under-studied and difficult to access
populations. Am. J. Ind. Med. 45:297–304, 2004. ß 2004 Wiley-Liss, Inc.

KEY WORDS: recruitment; construction safety; safety training; intervention

research

1
College of Nursing, Arizona State University,Tempe, Arizona
2
College of Nursing, University of New Mexico, Albuquerque, New Mexico
3
College of Medicine, Upstate Medical University, Syracuse, New York INTRODUCTION
4
National Institute for Occupational Safety and Health (NIOSH), Morgantown,
West Virginia, formerly of the Kentucky Injury Prevention and Research Center.
The views in this document are those of the authors and not necessarily those of the Safety training researchers face recruitment challenges
Upstate Medical University, Arizona State University, University of New Mexico, University when potential subjects are difficult to locate, engage, or
of Kentucky, National Institute of Occupational Safety and Health, or the United States Gov- sustain in training and research. Failure to recruit the required
ernment.The study that supported the preparation of this manuscript was conducted at the
Kentucky Injury Prevention and Research Center, University of Kentucky, Lexington, KY. sample size threatens scientific integrity and increases the
Dr. P. Kidd is deceased. costs per participant [Rosenberg et al., 1996]. High rates of
Contract grant sponsor: National Institute for Occupational Safety and Health; Contract refusal to participate in worksite health promotion studies
grant number: RO1/CCR413067.
*Correspondence to: Mark B. Parshall, College of Nursing, Mailstop MSC09 5350, have been reported in construction [Thompson et al., 1997].
1University of New Mexico, Albuquerque, NM 87131. E-mail: mparshall@salud.unm.edu Nationally, over 80% of construction companies have
ten or fewer employees [Ringen et al., 1995b]. Small con-
Accepted 28 October 2003
DOI 10.1002/ajim.10335. Published online in Wiley InterScience struction companies reportedly have higher injury severity
(www.interscience.wiley.com) rates as measured in fatalities, lost work time, injury-related

2004 Wiley-Liss, Inc.

298 Kidd et al.
worker turnover, and decreased availability of safety training
programs and occupational health services than larger con-
struction companies [Oleinick et al., 1995].
Construction workers in small companies are difficult to
access for several reasons. Small construction companies are
extremely dependent upon local economic conditions, hence,
they appear and disappear quickly. The nature and the site
of work change frequently, making it difficult to locate,
recruit, and retain research participants. Businesses are often
operated out of the owner’s home. Frequently, the only
‘‘company’’ phone may be a residential listing. When times
are bad, a commercial phone line, if any, may be considered a
dispensable luxury that can be disconnected. Jobs often come
through word-of-mouth or existing relationships with larger
contractors. Therefore, many small contractors do not
advertise in commercial phone directories. When work is
plentiful, the owner of a small construction company
(especially those in heavy or special trades) may have
regular work through a particular general contractor who
knows how and where to contact preferred subcontractors. At
the worker level, many who work for small companies do not
belong to a trade union, and it is not uncommon for a con-
struction worker to be employed by several small contractors
in a given year.
Even if one is able to access owners or workers, it may be
difficult to persuade them to participate in safety training
activities. Due to competition, narrow profit margins, and
ever-increasing operational costs, owners of small construc-
tion companies may feel that anything not directly related to
a specific job, and keeping work on schedule, may detract
from productivity [Blumenthal et al., 1995]. Hence, injury
prevention activities may also be viewed as a luxury, not a
necessity.
Preventive interventions are based on future risk for only
some proportion of a group [Stein et al., 1991; Capaldi et al.,
1997]. This makes it difficult to convey accurate information
about individual risk to construction workers. This, in turn,
may lead to a low level of motivation to participate in safety
interventions [Hansen et al., 1985]. Explicit acknowledg-
ment of high risk may produce negative mental health or
productivity consequences compared with the status quo.
Construction work, particularly for new, small companies,
may be so intermittent and insecure that the work itself
produces mental stress. This stress may impede healthful
behavior and participation in safety training [Ringen et al.,
1995b]. A higher value may be placed on services offering
immediate relief or concrete help [Ballard et al., 1993;
LaMontagne and Needleman, 1996].
Although, strictly speaking, occupational safety and
health statutes and regulations apply to all employers, as a
practical matter, companies below some size threshold
generally escape scrutiny. Hence, for small construction
owners and workers, research may be viewed with suspicion
as a ‘‘foot in the door’’ for unwanted regulatory oversight or
as an attempt to find violations of existing laws or rules.
Owners of small companies in some construction trades,
notably roofing, may feel that they already are at a dis-
advantage in terms of having been assigned higher workers’
compensation insurance multipliers than other trades. They
may fear continued or worsening disadvantage if participa-
tion in safety training and evaluation research leads to
increased scrutiny of their injury experience or existing
safety practices.
The purposes of this paper are to illustrate difficulties we
encountered in recruitment and to share lessons learned from
conducting a safety training and evaluation study with small
construction companies. Recruiting strategies were based on
recommendations in the literature on survey and interven-
tional research. We found that, in spite of following such
recommendations, recruitment and retention were not very
successful. Therefore, while continuing the study as planned,
we added questions pertaining to reasons for participation
or nonparticipation. Responses to these questions revealed
several major issues that were not prominent in literature on
recruitment. We base conclusions on our experiences with
this study and, to the extent possible, alternative recruitment
tactics for which there is some support in the literature.
MATERIALS AND METHODS
Design
Six narrative simulation exercises for small construction
companies using a quasi-experimental design were devel-
oped and tested. The intervention consisted of three narrative
simulation exercises targeting back injury prevention (year
1) or fall prevention (year 2). The study was approved by the
University’s Institutional Review Board. Informed consent
was obtained from all intervention participants. For the
control group, anonymous return of study measures con-
stituted consent.
Sample
The sample consisted of owner-operators and super-
visory and nonsupervisory employees of small construction
companies. To reach them, we formed a partnership be-
tween the Kentucky Injury Prevention and Research Center
(KIPRC) and Kentucky Employers’ Mutual Insurance
(KEMI). KEMI is the state fund workers’ compensation
insurer for Kentucky. It is, therefore, the largest workers’
compensation carrier for small business operations in the
state. This partnership permitted us to identify and recruit
small construction companies statewide for participation in
the intervention. At the same time, however, the partnership
gave rise to unintended constraints on study design and sam-
pling. First, because business owners are the policy holders,
policies can be used to locate owners, but not employees.
 Overcoming Recruitment Challenges in Construction
Thus, our ability to recruit employees was dependent on our
success in recruiting owners to participate in the safety
intervention. For similar reasons, to constitute a no-treatment
control group, we were essentially limited to a sampling
frame of company owners.
Sampling procedures for the overall study are described
in greater detail elsewhere [Kidd et al., 2000; Wojcik et al.,
2003]. Because the number of employees may fluctuate in
small businesses, KEMI does not index companies by
number of employees, but by self-reported payroll. Identi-
fication of small construction companies was accomplished
via cross-indexing KEMI records for companies with
reported annual payroll of $10,000 with standard industrial
classification (SIC) codes for general, heavy, or special
construction trades. Briefly, companies were randomly selec-
ted from among KEMI policy holders with the appropriate
SIC codes and payroll.
In the first year of intervention testing, assignment of
companies to treatment or control status was quasi-random
because the two intervention administration modes were
being compared: individual home administration and group
sessions in different geographic regions of the state. Assign-
ment to group administration was based, in part, on geo-
graphic proximity to sites where groups were scheduled. A
nonintervention control group of owners was randomly
selected for purposes of demographic and pretest compar-
isons. As the study progressed, the number who actually
participated in the group administration sessions was
insufficient for adequate testing of the relative efficacy of
that mode of administration. Therefore, in the second
and final year of intervention testing, only home administra-
tion was used versus a nonintervention control group.
Assignment of companies to intervention or control in year
2 was random, with the exception that any individual
intervention participant from the previous year was invited
to re-enroll.
Recruitment Procedures
Intervention participants
In either year, companies in the intervention sampling
frame received written invitations to participate. Invitations
were on project letterhead that clearly identified the project as
a joint endeavor of KEMI and KIPRC (hence, the University
of Kentucky). Letters were signed by the Chief Executive
Officer of KEMI and the project principal investigators and
mailed to company owners. Because there was a participation
incentive, a 10% policy premium discount at next renewal,
companies had to be currently insured by KEMI. Project staff
attempted to follow-up on all mailed invitations with a
telephone contact to the company owner to further explain
the study and the participation incentive, to answer questions,
and to encourage participation.
299
In years 1 and 2, for companies assigned to home
administration, upon expression of interest in participation,
sufficient intervention packets were mailed for the company
owner and employees to complete. Intervention packets were
to be distributed by the company owner, completed at home
by the worker, and then mailed directly back to the research
team. This ensured that owners would be blind to employees’
responses and to which, if any, employees participated. In
year 1, for companies invited to group administration
sessions, the date, time, and location of the group session
was confirmed and owners were asked to estimate how many
from their company would attend. Reminder phone calls
were made within a few days of the scheduled session.
Regardless of the mode of administration, intervention
packets consisted of an invitation letter, consent form,
instructions for completing the simulations and study mea-
sures (demographic and safety climate questionnaires), and
the three simulation exercises with associated evaluation
questionnaires. Answer sheets for the simulations were
printed in invisible ink which participants ‘‘develop’’ with a
special (‘‘latent-image’’) marking pen that was included with
the packets. For home administration participants, prepaid
return postage materials also were included. It was estimated
that it would take approximately an hour and a half to
complete the intervention materials.
Intervention participants were asked to return their
names and addresses on index cards so that they could receive
follow-up materials. These included comprehensive answer
booklets with underlying rationales for correct and incorrect
answers. These were mailed out as soon as the completed
packets were returned. A delayed posttest with retrospective
pretest measure of safety climate was mailed to participants
approximately 4 months after the intervention.
Control group participants
In either year, all control group members were owner-
operators only. Participation was sought as for an anonymous
mail survey, with return of study questionnaires constituting
consent for the control group. This approach was necessary
because it was not feasible for KEMI to offer a premium
discount simply for returning a survey. Owners were sent the
invitation, demographic and safety climate questionnaires,
and prepaid return postage materials. All who received
control packets, were mailed a follow-up postcard 2 weeks
later reminding them to return the questionnaires if they had
not already done so and thanking them for their participation
if they had.
Reasons for nonparticipation
Because participation was less than desired in year 1, we
added a question about nonparticipation to the control survey.
In addition, we sent the control survey out anonymously to
300 Kidd et al.

company owners who had been contacted about participation
by phone and who declined to participate or who agreed by
phone to participate but from whose companies no returns
were received. We also added a question about reasons for
participation (because owners were blinded to whether or not
anyone from their company returned intervention materials;
hence, some of those who agreed by phone to participate may
have thought, in good faith, that their company actually had
participated). In year 2, this process was repeated with
owners who agreed to participate but from whose company
no returns were received.
RESULTS
Summary data on recruitment and participation in the
intervention or control groups are shown in Tables I and II,
respectively. Because attendance at the group sessions was
disappointing after eight sessions the group administration
arm of the study was closed.
For this reason, the most balanced comparison between
years is between home administration cohorts. For home
administration, a higher percentage of owners contacted by
phone in year 2 agreed, in principle, to participate compared
with year 1 (56.7 vs. 36.2%, 95% CI for difference, 13.8,
27.2%). However, among those who agreed to participate by
phone, actual participation by companies (i.e., those from
whom at least one completed intervention packet was
returned) was approximately equal in each year (32.0% in
year 1 compared with 30.0% in year 2, ns). As a proportion of
the total number of intervention packets requested for home
administration in either year, there was no difference in
the percentage of individual returns by year (25.6% in year
1 compared with 23.2% in year 2, ns).
In the control group (Table II), there was no difference
between years in the percentage of participation among
owners who received study materials (15.2 vs. 12.7% in
years 1 and 2, respectively, ns). The difference between
aggregated home administration (122/397, 30.7%) and
control participation rates (103/737, 13.9%) was significant
(absolute difference, 16.8%, 95% CI for difference, 11.9,
21.7%), which, presumably, reflects combined influences of
direct phone contact during recruitment and the participation
incentive.
Reasons for Nonparticipation
and for Participation
Because it was necessary to recruit participation via
owners, and because of the owners-only composition of
control groups in either year, analyses of reasons against or
for participation was limited to owners. The most frequently
reported reasons for not participating in years 1 and 2 were:
‘‘time involved too much for me’’ (52% in year 1, 47% in year
2) and ‘‘company already has good safety record’’ (52% in
year 1, 60% in year 2). The response category rated ‘‘most
important’’ was, ‘‘time involved too much for me’’ (40% in
both years). Hardly any respondents (n ¼ 2) indicated
any concern that ‘‘the incentive was inadequate’’ or that
the intervention materials did not appear ‘‘interesting’’ or
‘‘beneficial.’’
In year 2, we also surveyed to discover reasons for
participation. The most frequently selected reasons for
participating were ‘‘will get insurance discount’’ (80%) and
‘‘safety is a priority for our company’’ (68%). Evidently, the
belief that a company’s safety record was already good could
be used by owners to justify either participation or
nonparticipation.
DISCUSSION AND CONCLUSIONS
Recruitment and participation rates were disappointing,
despite serial recruitment contacts and an incentive. Never-

TABLE I. Recruitment of Intervention Participants by Company, Individual,Year, and Mode of Participation; Small Construction Companies, Kentucky

Companies Individuals

Number Number of (%) owners Number of (%) returning Number of interventions Number (%) of interventions
invited by mail agreeing to participate intervention mailed out returned
Year1
Group 150 55 (36.7) 25 (45.5) NA 38 NA
Home 414 150 (36.2) 48 (32.0) 430 110 (25.6)
Subtotal 564 205 (36.3) 73 (35.6) 430 148 (34.4)
Year 2
Prior participants NA 73 (NA) 18 (24.7) 146 26 (17.8)
Home 436 247 (56.7) 74 (30.0) 573 133 (23.2)
Subtotal 436 320 (NA) 92 (28.8) 719 159 (22.1)
Total 1,000 525 (NA) 165 (31.4) 1,149 307 (26.7)

NA, not applicable; since interventions were administered on site.

 Overcoming Recruitment Challenges in Construction
TABLE II. Control Group Surveys Invitations and Returns; Small
Construction Companies, Kentucky
Number of (%) company Number (%) of company
owners invited by mail owners returning survey
Year1 382 58 (15.2)
Year 2 355 45 (12.7)
Total 737 103 (13.9)
theless, among companies contacted by phone for participa-
tion in the intervention group, the participation rate for home
administration was slightly more than double the return rate
for control group surveys. In essence, the approximately 17%
difference in participation rates between intervention
(31%) and control groups (14%) represents the combined
effect of the intervention recruitment strategies (primarily
telephone contact and the incentive) plus sampling error
(5%). Thus, it was not the case that the recruitment
strategies we used were utterly ineffective, they merely were
insufficient to garner satisfactory participation in this
difficult to access population.
We believed that our partnership with KEMI was critical
for obtaining statewide participation. KEMI enthusiastically
endorsed and supported the study. They provided the finan-
cial incentive on the insurance premium for companies that
participated in the intervention. Obviously, they could only
offer that incentive to companies they insured, and, for
economic reasons, they were not in a position to offer the
same incentive to companies in a no-treatment control group.
We acknowledge this as a weakness in study design, but, on
balance, we continue to believe the positive aspects of this
partnership outweighed the negative (e.g., without the
partnership, locating potentially eligible companies would
have been far more difficult). On the other hand, in hindsight,
we failed adequately to anticipate the consequences to study
design and, especially, sampling of having to recruit via
company owners. In future studies in this occupational
sector, investigators (and, for that matter, reviewers) would
do well to give greater attention at the proposal stage to
methods to enhance equivalence of control and intervention
groups. Despite the obvious virtues of random sampling in
that regard, it remains a methodological challenge to ge-
nerate random samples of sufficient size in a difficult to
access and locate population.
A further potential weakness of the incentive we offered
was that it rewarded owners but not necessarily employees.
Nevertheless, the participation and nonparticipation surveys
showed hardly any owners who felt the participation
incentive was inadequate, and, clearly, among those who
participated, the incentive was an important factor. Even so,
perceived time demands of participation (1½ hr) were viewed
as a barrier that, apparently, outweighed the incentive for
many potential participants.
301
Funding by the National Institute for Occupational
Safety and Health (NIOSH) may have been perceived as a
barrier to participation because many potential participants
incorrectly assume that NIOSH is administratively related to
the Occupational Health and Safety Administration (OSHA).
OSHA is viewed by small construction company owners and
workers as the bearer of rules, fines, and unwanted regulation.
Similar concerns about blurred boundaries between research
and enforcement activities pertaining to risk abatement
have been described in both community and occupational
settings [LaMontagne and Needleman, 1996; Needleman
and Needleman, 1998].
Being endorsed by the workers’ compensation insurance
carrier may have also had negative repercussions with respect
to recruitment. For example, any company owner who ever
misrepresented any information to the insurer (e.g., size of
payroll) or failed to report a claim, might have feared being
‘‘found out,’’ despite assurances that only summary data
would be reported.
It also is noteworthy that the owner’s belief about the
existing safety record and practices of their companies was
invoked as both a barrier against and as a motivator for
participation, arguably an issue of perceiving a glass as half-
empty versus half-full. In essence, owners’ initial judgments
about time demands of participation appear to have been far
more decisive than their beliefs about their safety records.
Accordingly, it would appear that recruitment efforts in this
population might benefit from designing interventions to
minimize time demands combined with efforts to persuade
company owners that safety training is worthwhile regardless
of one’s past record. Given that there was some recruitment
advantage to mail plus telephone contact versus mail only,
increased contact with owners (e.g., face-to-face) might
further enhance recruitment.
The theory of social influence proposes that researcher
credibility and influence is related to similarity between
researcher and participant [Gauthier and Clarke, 1999]. A
match of values, beliefs, and attitudes between the two is
desired. Theoretically, having a contact person who is
knowledgeable about the field and can ‘‘talk shop’’ with
potential enrollees is another recruitment strategy. However,
in year 1, our recruiter was a part-time general contractor.
Even with that expertise, recruitment fell far short of
expectations. In year 2, a different individual, without a
background in construction, was the primary recruiter, with
little if any difference in recruitment rates. Accordingly, this
strategy, plausible though it sounds, may be helpful but may
not be sufficient for adequate recruitment.
We conclude with a number of lessons we learned from
conducting the study that might be useful when recruiting
small construction companies for participation in safety
training. When available, literature-based support is pro-
vided for the suggested strategies and framed within the
context of small construction companies. The strategies have
302 Kidd et al.
been organized in two areas: sampling considerations and
perspectives of participants.
Sampling Considerations
Do not rely exclusively on owners
to access workers
Decide at what level you will intervene, company versus
worker, and either directly access the appropriate target
group or use a strong referral mechanism. In our study, the
owner had to select workers to complete the intervention
packet and distribute the packets. The company was
contacted and recruited by the research team but the worker
was the focus of recruitment. This extremely hard-to-reach
subgroup may be best recruited through face-to-face contact
at the jobsite using a peer referral technique. Face-to-face
contact has increased recruitment in some prevention
oriented studies [Little et al., 1979; Capaldi et al., 1997],
but probably adds to the time and expense of a study.
Consider the trade-offs between
using representative versus
convenience sampling
Trade-offs between accessing companies using a
representative sampling frame, such as in this study, or
recruiting through referral and personal contact should be
weighed. Participation tends to be higher using the latter
method [Schlernitzauer et al., 1998].
Arguably, in this study, we might have attempted to
recruit companies through trade associations, unions, the
media, or by visiting job sites. However, regardless of size,
less than 20% of all construction companies belong to a trade
association or construction organization [Ringen et al.,
1995a]. In Kentucky, most small construction companies
are not unionized. Recruitment via newspapers or broadcast
media may be more feasible when recruitment is confined to a
limited number of localities. It seems likely that the same
would be true of job site visits, assuming sites can be located
reliably.
Obtaining support from general contractors and visiting
areas that they frequent is another suggestion. This may
involve contacts via material suppliers as well as identifying
places where business may be conducted informally (e.g.,
lumber and home improvement centers that deal frequently
with small contractors).
Oversample to obtain the needed
number of participants
In literature on clinical trials, it has been suggested that
the pool(s) from which participants are recruited should
exceed the desired sample size by a factor of 10 [Ederer,
1975]. Our experience suggests that in prevention trials,
such as ours, it may be necessary to build in some similar
multiplier.
Do not underestimate the amount
of time needed to contact and
recruit participants
We found that recruitment and retention of small
construction companies required far more time, effort, and
resources than we had anticipated. On average it took three
calls per company to successfully contact an owner. Phone
numbers change. Many people use answering machines to
screen calls. When it is a family business, calls may be
answered by someone other than the business owner. When
times are lean, phones may get disconnected. Even when
contact was established, many still declined to participate.
When developing the budget for the project, we under-
estimated these difficulties of recruitment. Hence, these were
not considered to be large enough expenses to justify a line
item. In retrospect, the budget should have reflected at least a
50% full-time equivalent position for recruitment efforts. In
addition, the timeline for the project should have allowed
ample time (e.g., 6–9 months) specifically to develop the
infrastructure needed to ensure adequate participation.
Assess stability of your
targeted sample
Because KEMI was a new company (less than 2 years in
operation at the initiation of our partnership), prior insurance
records were not available. If they had been available, we
could have screened for companies that had remained with
KEMI at least 2 years indicating some type of stability. The
likelihood that companies struggling to keep afloat financially
would value safety sufficiently to participate in a safety-
training program is probably slim. Yet we had no means
whereby such companies could be identified or screened out.
Perspective of Participants
Examine respondent burden
The degree of perceived intrusiveness also influences a
person’s willingness to participate in a study [Rosenberg
et al., 1996]. Home completion of the safety-training
program may have been preferable to group administration,
but it still required approximately an hour and a half. Thus, it
may have intruded on participants’ family time or other
commitments. Also, there was no mechanism for ensuring
employees were paid for time spent completing the study
materials. Given the competitive and economic pressures
small companies face, our experiences suggest that, when in
doubt about time commitments, potential participants either
 Overcoming Recruitment Challenges in Construction
decline out of hand or may judge participation as too
burdensome even after agreeing in principle to participate.
Accordingly, initial contacts must emphasize how potential
benefits will be worth the time entailed in the participation.
Another demand placed on the respondent may have
been the reading required to complete the simulation
exercises. Although written at a sixth-grade level, degree of
literacy was not assessed in the sample selection process and
may have contributed to a lower participation rate.
Pilot recruitment strategies
In retrospect, many of our assumptions about recruiting
were naı̈ve. Ideally, had we piloted our recruitment strategies,
it might have allowed time to identify and revise problem areas
before actual study implementation. A potential alternative to
a pilot recruitment study would be to conduct focus groups or
an ethnographic pilot study to assess the relevance of the study
from potential participants’ perspectives and to solicit their
ideas for motivating participation. Prior to the intervention
phase, we did conduct focus groups to develop the simulations
and to critique their content. Had we anticipated recruitment
difficulties, it would have been a relatively simple matter to
have added to the interview guide questions about possible
recruitment strategies, or to have conducted several additional
focus groups specifically to solicit ideas that might benefit
recruiting with this population.
Tailor incentives to the target group
and minimize disincentives
Incentives have been linked with recruitment in worksite
health promotion programs [Jeffrey et al., 1993]. There is wide
variability in the effects of incentives across groups in pro-
moting recruitment [Hoffman et al., 1998]. Types of incentives
that are appropriate may vary based on the characteristics of
the target group or the degree of inconvenience associated with
participation [Pavlik et al., 1996]. We were unable to offer
direct incentives to workers, due to budgetary constraints.
Cash or a gift certificate might well have been more mean-
ingful to them than an insurance discount for the company.
Studies have shown that an incentive given at time of
recruitment increased participation [Wiseman et al., 1983]
while rewards promised after participation did not [Cook
et al., 1985]. In our study, the insurance premium discount
was intended to be an incentive. However, because of the
nature of the discount, it functioned more like a follow-up
reward. In addition, the reward went to the company, not to
the individual completing the training. These discounts were
realized at the time of policy renewal, which may have
been up to a year from participation in the study (e.g., if a
company had just renewed its policy immediately prior to
participation). Other disincentives may have included own-
ers’ concerns about past underreporting of payroll or injuries,
303
or letting a foot in the door for unwanted regulatory scrutiny.
We do not claim that it is possible to eliminate all such
perceived disincentives, but greater attention to potential
disincentives when a study is designed might help minimize
some perceived barriers to participation.
SUMMARY
The recruiting of small construction companies in safety
training and evaluation research presents a challenge for
investigators. The injury and fatality rates in the construction
industry and among small construction companies provides
the rationale for accepting this recruitment challenge. Our
experiences suggest that, despite concerted efforts, conven-
tional approaches to recruitment were less than effective in
reaching this population. In particular, we found much greater
difficulty than anticipated in contacting company owners by
phone to explain the study in detail, and we had to rely on them
to motivate workers to participate. Although the incentive we
offered appeared to be adequate to owners, there was no clear
subsidiary incentive for employees. Moreover, we failed to
appreciate the degree to which owners’ initial judgments about
time demands of participation might outweigh the potential
benefits of a discounted insurance premium. Because the
study was not designed as a formal test of alternative
recruitment strategies, our suggestions for alternative strate-
gies are untested in this population. Nevertheless, we hope that
analyzing our overall lack of success and attempting to
articulate various lessons learned will prove useful to other
investigators contemplating safety intervention studies in this
or other difficult to reach populations.
ACKNOWLEDGMENTS
The lead author of this article, Dr. Pamela Kidd, died
suddenly in December, 2002. We dedicate this article to the
memory of our beloved colleague, mentor, and friend.
Revision of the article was completed by the co-authors. The
authors also would like to acknowledge the financial support
from the National Institute for Occupational Safety and
Health (NIOSH) and the technical assistance of Dr. Mike
Colligan of NIOSH throughout the project. We also
acknowledge the generous support of Kentucky Employers
Mutual Insurance (KEMI); in particular, we thank Roger
Fries, Robert Jameson, and Jonathan Mays of KEMI for their
enthusiasm and support for the project.
REFERENCES
Ballard EL, Nash F, Raiford K, Harrell LE. 1993. Recruitment of black
elderly for clinical research studies of dementia: The CERAD
experience. Gerontologist 33:561–565.
Blumenthal DS, Sung J, Coates R, Williams J, Liff J. 1995. Recruitment
and retention of subjects for a longitudinal cancer prevention study in an
inner-city black community. Health Ser Res 30:197–205.
304 Kidd et al.
Capaldi DM, Chamberlain P, Fetrow R, Wilson JE. 1997. Conducting
ecologically valid prevention research: Recruiting and retaining a
‘‘whole village’’ in multi-method, multi-agent studies. Am J Commun
Psychol 25:471–492.
Cook JR, Schoeps N, Kim S. 1985. Program responses to mail surveys as
a function of monetary incentives. Psychol Rep 57:366.
Ederer F. 1975. Practical problems in collaborative clinical trials. Am
J Epidemiol 120:111–118.
Gauthier MA, Clarke WP. 1999. Gaining and sustaining minority
participation in longitudinal research projects. Alzheimer Dis Assoc
Disord 13:S29–S33.
Hansen WB, Collins LM, Malotte CA, Johnson CA, Fielding JE. 1985.
Attrition in prevention research. J Behav Med 89:261–275.
Hoffman SC, Burke AE, Helzlsouer KJ, Comstock GW. 1998.
Controlled trial of the effect of length, incentives, and follow-up
techniques on a response to a mailed questionnaire. Am J Epidemiol
148:1007–1011.
Jeffrey RW, Forster JL, Baxter JE, Kelder SH. 1993. Association for
Worksite Health Promotion. Am J Health Promot 8:98–100.
Kidd PS, Struttmann T, Wojcik S, Parshall MB. 2000. Loss education to
reduce construction-related injuries. Final Report, DHHS, PHS, CDC/
NIOSH: R01/CCR413067.
LaMontagne AD, Needleman C. 1996. Overcoming practical chal-
lenges in intervention research in occupational health and safety. Am
J Ind Med 29:367–372.
Little RE, Streissguth AP, Page EL. 1979. Techniques for recruiting
special types of persons for research: Pitfalls and successes in enlisting
recovered alcoholic women. Public Health Rep 94:332–335.
Needleman CE, Needleman ML. 1998. Science versus service in a
community-based lead screening: Reasons for low participation. In:
Leviton LC, Needleman CE, Shapiro MA, editors. Confronting public
health risks: A decision-maker’s guide. Thousand Oaks, CA: Sage,
p 58–88.
Oleinick S, Gluck JV, Guire KE. 1995. Establishment size and risk of
occupational injury. Am J Ind Med 28:1–21.
Pavlik VN, Hyman DJ, Vallbona C, Dunn JK, Louis K, Dewey CM,
Wieck L, Toronjo C. 1996. Response rates to random digit dialing for
recruiting participants to an onsite health study. Public Health Rep 111:
441–445.
Ringen K, Englund A, Welch L, Weeks JL, Seegal JL. 1995a.
Perspective on the future. Occup Med 10:445–451.
Ringen K, Seegal JL, Englund A. 1995b. Safety and health in the
construction industry. Ann Rev Public Health 16:165–188.
Rosenberg R, Gagnon M, Murphy-Gismondi P, Lock Ooi W, Kiel
DP, Lipsitz LA. 1996. Factors influencing subject willingness to
participate in clinical gerontologic research. Aging Clin Exp Res 8:
400–408.
Schlernitzauer M, Bierhals AJ, Geary MD, Prigerson HG, Stack JA,
Miller MD. 1998. Recruitment methods for intervention research in
bereavement-related depression. Am J Geriatr Psychiatry 6:67–74.
Stein RE, Bauman LJ, Ireys HT. 1991. Who enrolls in prevention trials?
Discordance in perception of risk by professionals and participants. Am
J Commun Psychol 19:603–617.
Thompson B, Leynseele J, Beresford SA. 1997. Recruiting worksites
to participate in a health promotion study. Am J Health Promotion 11:
344–351.
Wiseman F, Schafer M, Schafer R. 1983. An experimental test of the
effects of a monetary incentive on cooperation rates and data collection
costs in central-location interviewing. J Market Res 20:439–442.
Wojcik S, Kidd P, Parshall M, Struttmann T. 2003. Performance and
evaluation analysis of small construction safety training simulations.
Occup Med 53:279–286.