Content uploaded by Melissa A Polusny
Author content
All content in this area was uploaded by Melissa A Polusny on Aug 18, 2014
Content may be subject to copyright.
R E S E A R C H A R T I C L E Open Access
Impact of different privacy conditions and
incentives on survey response rate, participant
representativeness, and disclosure of sensitive
information: a randomized controlled trial
Maureen Murdoch
1,2,3*
, Alisha Baines Simon
2
, Melissa Anderson Polusny
2,4,5
, Ann Kay Bangerter
2
,
Joseph Patrick Grill
2
, Siamak Noorbaloochi
2,3
and Melissa Ruth Partin
2,3
Abstract
Background: Anonymous survey methods appear to promote greater disclosure of sensitive or stigmatizing
information compared to non-anonymous methods. Higher disclosure rates have traditionally been interpreted
as being more accurate than lower rates. We examined the impact of 3 increasingly private mailed survey
conditions—ranging from potentially identifiable to completely anonymous—on survey response and on
respondents’representativeness of the underlying sampling frame, completeness in answering sensitive survey items,
and disclosure of sensitive information. We also examined the impact of 2 incentives ($10 versus $20) on these outcomes.
Methods: A 3X2 factorial, randomized controlled trial of 324 representatively selected, male Gulf War I era veterans who
had applied for United States Department of Veterans Affairs (VA) disability benefits. Men were asked about past sexual
assault experiences, childhood abuse, combat, other traumas, mental health symptoms, and sexual orientation. We used a
novel technique, the pre-merged questionnaire, to link anonymous responses to administrative data.
Results: Response rates ranged from 56.0% to 63.3% across privacy conditions (p= 0.49) and from 52.8% to 68.1% across
incentives (p= 0.007). Respondents’characteristics differed by privacy and by incentive assignments, with completely
anonymous respondents and $20 respondents appearing least different from their non-respondent counterparts.
Survey completeness did not differ by privacy or by incentive. No clear pattern of disclosing sensitive
information by privacy condition or by incentive emerged. For example, although all respondents came from the same
sampling frame, estimates of sexual abuse ranged from 13.6% to 33.3% across privacy conditions, with the highest
estimate coming from the intermediate privacy condition (p=0.007).
Conclusion: Greater privacy and larger incentives do not necessarily result in higher disclosure rates of sensitive
information than lesser privacy and lower incentives. Furthermore, disclosure of sensitive or stigmatizing information
under differing privacy conditions may have less to do with promoting or impeding participants’“honesty”or
“accuracy”than with selectively recruiting or attracting subpopulations that are higher or lower in such experiences.
Pre-merged questionnaires bypassed many historical limitations of anonymous surveys and hold promise for exploring
non-response issues in future research.
Keywords: Randomized trial, Patient surveys, Participation bias, Non-response bias, Anonymity, Confidentiality
* Correspondence: Maureen.Murdoch@va.gov
1
Section of General Internal Medicine, Minneapolis VA Medical Center,
Minneapolis, MN, USA
2
Center for Chronic Disease Outcomes Research, Minneapolis VA Medical,
One Veterans Drive, Minneapolis, MN 55417, USA
Full list of author information is available at the end of the article
© 2014 Murdoch et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain
Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article,
unless otherwise stated.
Murdoch et al. BMC Medical Research Methodology 2014, 14:90
http://www.biomedcentral.com/1471-2288/14/90
Background
Surveys represent one of the most efficient and inexpen-
sive research methods available to collect representative,
high quality data from large numbers of research partici-
pants. They therefore frequently serve as the backbone
used to define the scope and magnitude of many poten-
tial public health problems. In the United States, for ex-
ample, large national surveys have been used to estimate
what proved at the time to be surprisingly high levels of
mental illness within the general population [1], physical
violence within families [2], and sexual assault among
women [3]. Even the United States Census, which serves
as the basis of apportioning Congressional representatives
and taxes to each state, is survey-based. Typically, survey
data are either collected by interviewers using face-to-face
or telephone communication with the participant or via
the participant’s own self-report.
Regardless of the topic studied and how the information
is collected, scientifically correct, survey-based prevalence
estimates require that research participants be representa-
tive of the population from which they are drawn, that
participants actually answer the questions that are asked
of them, and that they answer those questions honestly.
On average, research participants disclose sensitive and
personal information, such as mental health symptoms,
drug misuse, and history of sexual assault more fre-
quently when responding to self-administered question-
naires than when taking part in face-to-face or telephone
interviews [4-7]. Studies suggest that disclosure of sensi-
tive information on self-administered questionnaires is
enhanced yet more when participants respond anonym-
ously instead of confidentially [5,8-11]. This implies
that anonymous, self-administered surveys may be the
optimal method for accurately cataloging information
about certain public health problems, such as the preva-
lence of physical or sexual abuse or of mental health
symptoms.
Although by no means proven, most survey researchers
take the stance that methods that generate higher preva-
lence estimates for stigmatizing or sensitive information
are probably more accurate than methods that generate
lower estimates. This stance, however, rests upon a rather
unlikely assumption that all people carry the same pro-
pensity to participate in survey research. Particularly when
a survey topic is sensitive, survey respondents tend to dif-
fer substantially from non-respondents [12]. Therefore,
three mechanisms might explain why anonymous surveys
generate higher prevalence estimates of stigmatizing or
sensitive information compared to non-anonymous sur-
veys: 1) propensity to participate in research is in fact
equal across all members of a sampling frame, and an-
onymous methods promote more honest self-disclosure
among the participants with stigmatizing experiences;
2) sampling frame members with stigmatizing experiences
are more reluctant than others to participate in surveys,
but anonymous methods reduce this inherent reluc-
tance (under selection is reduced); 3) anonymous methods
disproportionately increase the propensity of people with
stigmatizing experiences to participate in the survey rela-
tive to those without such experiences (over selection
is induced). The first two mechanisms reduce bias; the
last introduces bias. Without information about non-
respondents’characteristics relative to respondents’,how-
ever, one cannot determine which possibility is correct.
Unfortunately, under typical anonymous conditions, such
information is unavailable.
Anonymous surveys carry other drawbacks relative to
confidential surveys. For example, unlike confidential
survey methods, anonymous survey responses cannot be
linked to administrative or other non-survey data, thus
limiting anonymous data’s richness and utility. Also, un-
less creative methods are employed, researchers often can-
not track or send follow-up mailings to non-respondents
of anonymous surveys, thus obtaining inferior response
rates (e.g., [13]). While low response rates do not neces-
sarily correlate to poor data quality, risks for non-response
bias do increase with lower response rates.
Two methods to bypass the tracking limitation in an-
onymous surveys have been described. In one, partici-
pants return a completed survey and a separately mailed
postcard. Only the postcard contains a unique identifier,
which is used to track respondents [14-16]. However,
this method increases respondent burden, which can re-
duce response rates. Furthermore, participants may find it
confusing and hence return only one item –e.g., the sur-
vey or the postcard, but not both. Receipt of equal num-
bers of postcards and surveys do not necessarily mean the
same people returned both. Even when both are returned
by the same person, the survey may be received consider-
ably earlier than the postcard. The participant may there-
fore be subjected to additional mailings until the postcard
is received, which may be annoying, and the researcher
may incur unnecessary mailing expenses. Finally, unbe-
knownst to the researcher, some respondents may return
more than one survey, leading to the overweighting of
those individuals’responses.
A second approach uses tracking envelopes, which sim-
plifies respondent burden, circumvents the problem of
postcards and surveys returning at different times, and
avoids analyzing multiple responses from a single partici-
pant [17]. In this approach, the envelope contains a unique
identifier, but not the survey. The two are returned to-
gether but separated immediately upon opening. Received
surveys are then intermixed in some random fashion to
avoid any possibility of linking them back to their ori-
ginal envelopes. If one participant returns more than
one envelope-survey pair, all but the first is discarded.
Until the envelope and survey are separated, however,
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 2 of 11
http://www.biomedcentral.com/1471-2288/14/90
thesurveyisnottrulyanonymous.Participantsmust
rely on the researcher’s integrity to maintain anonymity,
and they may be less willing to disclose sensitive informa-
tion relative to the postcard tracking method, where priv-
acy is absolute. Each approach has pros and cons, but the
two’s effect on response rates, survey completeness, or dis-
closure of sensitive information have never been directly
compared.
In the present paper we address these issues using a
novel technique we developed, the pre-merged question-
naire, which allows comparisons between respondents
and non-respondents even under anonymous survey
conditions. The study involved a potentially sensitive,
self-administered questionnaire asking about several
traumatic experiences, including sexual assault during
military service. The population of interest was male US
Gulf War I era veterans with possible posttraumatic
stress disorder (PTSD) who had previously applied for
Department of Veterans Affairs (VA) disability benefits.
We had reason to believe that sexual assault experiences
were particularly high in this population [18]. However,
we also feared that traditional rape myth beliefs [19],
which may be especially strongly held by military ser-
vice members socialized into a masculinized subcul-
ture, might either deter male sexual assault survivors’
participation in the research or impede their disclos-
ing of such experiences.
Using 3 levels of increasing privacy tied to the tracking
methods described above, we hypothesized that re-
sponse rate and participant representativeness, the
number of sensitive questions actually answered by
participants, and the proportion of participants disclos-
ing potentially sensitive information would increase in
a dose-response manner from the lowest to highest priv-
acy condition. Because higher incentives consistently
improve survey response [20], we also tested the im-
pact of two incentives, $10 versus $20, on survey re-
sponse. We expected the response rate, number of
sensitive questions answered, and proportion of par-
ticipants disclosing sensitive information would be
higher among those receiving the $20 incentive com-
pared to the $10 incentive.
Methods
Population and setting
We used simple randomization without replacement to
select 324 veterans for survey from the population of
46,824 men who applied for VA PTSD disability benefits
prior to June 2007 and had served in the US Armed
Forces between August 2, 1990 and July 31, 1991.
Study design and assignment
The study was a 3X2 factorial, randomized controlled
trial (Figure 1). Using simple randomization, Veterans
were assigned to one of 3 tracking/privacy condi-
tions:
1) “Confidential”: Under the least private condition,
veterans received a survey with a highly visible,
coded, unique identifier affixed to the front page of
the survey. This was used for tracking, and
individual respondents were potentially identifiable
from their surveys.
2) “Anonymized-Envelope”: Intermediate in privacy,
veterans were asked to return their surveys in a
study envelope, which had a pre-printed, unique
identification number (ID) on it. When the
completed questionnaire was returned, study
personnel immediately separated it from the
envelope. The questionnaire was intermixed with
other arriving surveys and set aside. The envelope
ID was used to indicate who had returned surveys.
Technically, as long as the survey resided within the
envelope, respondents could be identified. Thus,
this method was not fully anonymous. Once the
questionnaire was removed from the envelope,
however, there was no longer any way to identify the
respondent (hence the term “anonymized”).
3) “Anonymous-Postcard”: The most private condition,
veterans returned their surveys in unmarked
envelopes. Besides the survey, veterans were also
asked to return an enclosed, brightly colored
postcard, which had a unique ID to allow tracking.
Respondents could not be identified from their
surveys or envelopes at any time.
Once Veterans were assigned to their tracking/privacy
condition, we then used simple randomization within each
condition to assign them to receive $10 or $20.
Protocol
Data collection
For all groups, the initial mailing included a cover letter
describing the study’s risks and benefits, the cash incen-
tive, and 25-page questionnaire. At two week intervals,
non-respondents were mailed a post-card reminder, a
second mailing of the survey, and a final mailing of the
survey via overnight mail (Federal Express). Cover letters
were printed on Minneapolis VA Medical Center letter-
head and listed the study’s funding agency. Veterans
were told that they had been selected for survey because
they had filed a VA disability claim and had served dur-
ing Gulf War I. They were also told that the survey
would ask about “combat, unwanted sexual attention,
and other lifetime and military experiences”. The cover
letters also stated in bold-face font, “We would like to
hear from you even if you never experienced combat or
unwanted sexual attention. We would also like to hear
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 3 of 11
http://www.biomedcentral.com/1471-2288/14/90
from you even if you were not deployed to the Persian
Gulf.”Cover letters were the same across groups, except
that they described the incentive, tracking method, and
privacy protections that were specific to each group.
Copies of cover letters are available upon request.
Pre-merged questionnaires
To our knowledge, we are the first to develop pre-merged
questionnaires for use in anonymous surveys. However,
pre-merged questionnaires are simply an extension of the
common strategy of using different colored paper, say, to
collect data from different groups (e.g., green paper for
men, yellow for women). Instead of different colored pa-
pers, however, we created a sticker that was applied to each
subject’s questionnaire just before mailing. The sticker was
designed to be as unobtrusive as possible and was thus
camouflaged as a return address on the survey’s back page
(Figure 2). Just below the address, we embedded an alpha-
numeric code into the mailcode, which corresponded to
key administrative data associated with each potential
subject. When the survey was returned, so was the admin-
istrative data—already merged. The sticker code was delib-
erately intended to be non-exclusive to the subject. For
example, a code such as “504ADBY”, indicating a veteran
was aged 50 years or older, served 4 years in the Army and
received disability benefits from the VA, could apply to
hundreds of thousands of veterans.
We maintained two separate, but interrelated computer-
ized files: an administrative file containing subjects’name
and administrative codes, which were used to generate the
stickers, and a tracking file containing their names and
tracking ID. As envelopes, postcards, or confidential sur-
veys were returned, the tracking ID was entered into the
tracking file. This action deleted respondents’name and ID
from the tracking file and triggered a simultaneous deletion
of their name and administrative code from the adminis-
trative file. Thus, by study’s end, only non-respondents’
administrative codes remained in the computerized rec-
ord. These were then used to compare non-respondents
to respondents. Respondents’administrative codes were
Figure 1 Study flow chart.
Figure 2 Example of a pre-merged sticker. For the present study, the sticker was placed within a pre-printed box on the last page of the survey.
In this example, administrative data begins after the “E”in the Mailcode.
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 4 of 11
http://www.biomedcentral.com/1471-2288/14/90
recaptured from the sticker on their returned question-
naires and hand-entered back into the analytical frame.
Measures
Primary outcomes
The primary outcome was response rate, calculated as
the number of returned surveys divided by the number
of veterans assigned to each arm.
Secondary outcomes
Secondary outcomes included the representativeness of
respondents, percentage of Veterans fully completing all
sensitive survey items, and the percentage disclosing sen-
sitive information. Information collected by the survey
that we thought might be sensitive included veterans’ex-
periences of sexual abuse, including sexual assault during
the time of Gulf War I; other traumatic experiences, in-
cluding combat and childhood physical abuse; mental
health problems, including depression, PTSD, and prob-
lem drinking; and veterans’sexual orientation.
Representativeness of respondents We used data from
the pre-merged sticker to compare respondents to non-
respondents. Available data included age greater than or
equal to 50 years versus younger, service in the Army
versus other branch, VA disability benefit status (receiv-
ing versus not), and any VA health care utilization versus
none. Specifically, we assessed whether the participant
had made a visit to any VA medical facility in the prior
year for any reason or had made visits to a VA facility
for primary or mental health care. The term, original
sample, refers to all veterans selected for the survey, re-
gardless of their response status. Responders and respon-
dents refer to the subset of veterans from the original
sample who returned surveys, and non-responders/non-
respondents refer to the subset of veterans who did not
return surveys.
Sensitive information Sensitive information was col-
lected by the survey and included the following:
Sexual abuse
We used 3 items from Sexual Harassment Inventory’s
criminal sexual misconduct scale [21] plus one
additional item [22] to assess sexual assault during the
time of Gulf War I, 4 items from the Sexual Abuse
subscale of the Childhood Trauma Questionnaire [23]
to assess childhood sexual abuse, and one item from
the Life Stressor Checklist [24] to assess any sexual
assault in the past year. A positive response to any one
of these questions indicated a history of sexual abuse.
Other traumatic experiences
Other traumatic experiences included combat
exposure, assessed using an adapted Combat Exposure
Inventory [25] version; childhood physical abuse,
assessed using items from the Childhood Trauma
Questionnaire’s relevant subscale [23]; and past-year
traumas, assessed using an adaptation of the Life Stressor
Checklist [24]. Veterans who reported any childhood
physical abuse item more than “rarely”were considered
physically abused.
Mental health problems
We used the 5-item RAND Mental Health Battery [26]
to assess depression, the Penn Inventory for PTSD [27]
to assess PTSD symptoms, and the TWEAK [28]to
assess alcohol misuse.
Sexual orientation
Sexual orientation was assessed using a single survey
item that read, “People are different in their sexual
attraction to other people. Which best describes your
feelings?”Responses ranged from 1=“Completely
heterosexual or ‘straight’” to 5=“Completely
homosexual or ‘gay’”. Responses were dichotomized as
“Completely heterosexual”versus “Not completely
heterosexual”for analysis.
Power
The study was funded to examine different incentives’im-
pact on response rate and had 80% power to detect a 10%
difference in response rates across incentives, assuming a
60% response rate in the $10 group and two-tailed alpha
of 0.05.
Analysis
The study was intended to examine main effects, but in-
teractions were assessed in an exploratory fashion. Results
are reported for tracking/privacy condition first; incentive
condition second; and, when tested, interactions third. We
used χ
2
tests to compare outcomes across privacy condi-
tions and incentives and to compare respondents and
non-respondents. We used logistic regression to test for
interactions between tracking/privacy condition and
incentive on outcomes. We used IBM SPSS Statistics
(version 19) and SAS (version 9.2) statistical packages
for analyses.
Masking, disclosure, and ethical approval
Data collectors and analysts were aware of study group
assignment. The Minneapolis VA Medical Center’s
Subcommittee for Human Studies approved the protocol.
Results
Response rate
Response rate overall was 60.5% and did not differ signifi-
cantly across tracking/privacy assignments (Confidential
response rate = 56.0%, Anonymized-envelope response
rate = 63.3%, Anonymous postcard response rate = 62.3%,
p= 0.49). However, the response rate was almost 15 full
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 5 of 11
http://www.biomedcentral.com/1471-2288/14/90
percentage points higher among veterans randomized to
receive the $20 incentive (response rate = 68.1%) com-
pared to the $10 incentive (response rate = 52.8%, p=
0.007). While the lowest response rate was obtained from
men randomized to the Confidential/$10 incentive group
(response rate = 43.6%; see Figure 1), tests for interactions
between tracking/privacy and incentives on response rate
were not statistically significant (p=0.46).
Respondent representativeness
As Table 1 shows, randomization failed to evenly distribute
the 324 veterans according to their past-year VA health
care utilization. Specifically, veterans randomized to the
Anonymous-Postcard were less likely to have made a VA
health care visit of any kind in the past year than were
veterans randomized to the Anonymized-Envelope and
Confidential groups (67.6% versus 75.4% in the other two
conditions). Otherwise, randomization successfully distrib-
uted all the remaining administrative characteristics evenly
across all the tracking/privacy and incentive conditions.
The characteristics of survey responders are shown in
Table 2. Responders in the Anonymized-Envelope group
had a higher proportion of individuals aged 50 years or
older, a lower proportion of white persons, and a lower
proportion of persons working for pay compared to the
other two groups, but none of these differences were sta-
tistically significant (all p’s > 0.18). Consistent with the
original sample’s maldistribution, Anonymous-Postcard re-
spondents were less likely than other respondents to have
made a visit of any kind to a VA medical facility in the
prior year. Compared to the other tracking/privacy condi-
tions, Anonymous-Postcard respondents were also sub-
stantially less likely to have made a mental health care visit
to a VA medical facility, but this could not be attributed to
a maldistribution of the original sample. Compared to the
administrative record, all respondents substantially under-
reported receiving VA disability benefits.
Respondents in the $10 incentive arm were signifi-
cantly older, less likely to be working for pay, and more
likely to say they received VA disability benefits than re-
spondents in the $20 incentive arm. Both groups substan-
tially underreported their receipt of VA disability benefits
compared to the administrative record. There were no sta-
tistically significant tracking/privacy-by-incentive interac-
tions (all p’s > 0.20).
Table 3 presents information for the original sample,
stratified by response status and by study assignment. Find-
ings show that Confidential and Anonymized-Envelope re-
spondents differed significantly from their non-respondent
counterparts in terms of age and service branch. Compared
to their non-respondent counterparts, Confidential re-
spondents were also more likely to be receiving VA
disability benefits, and Anonymized-Envelope respon-
dents were more likely to have made VA primary care
and mental health visits. There were significant age differ-
ences among respondents and non-respondents random-
ized to receive $10, but respondents and non-respondents
did not differ significantly on any available characteristic
among those assigned to the Anonymous-Postcard or $20
incentive. There were no significant tracking/privacy-by-
incentive interactions.
Percentage fully completing sensitive items and percentage
disclosing sensitive information
As Table 4 shows, with the exception of combat items, re-
spondents answered every item on each of the potentially
sensitive scales more than 90% of the time, regardless of
tracking/privacy condition or incentive. Twenty-six ques-
tions were used to assess combat exposure, which may
explain why it had the most skipped items (10.7% over-
all), though respondents were twice as likely to skip a
combat item as they were to skip a PTSD item (3.1% over-
all), which also contained 26 questions. The sexual abuse
questions were second most likely to be skipped (7.1%
overall). There were no statistically significant associations
between tracking/privacy assignment and completion of
sensitive survey items. Likewise, higher incentive was not
associated with greater completion of sensitive survey
Table 1 Population characteristics by tracking/privacy condition and incentive; results reported as a percentage (%)
Characteristic Overall By tracking/privacy condition Incentive
Confidential Anonymized-Envelope Anonymous-Postcard $10 $20
N= 324 n= 109 n= 109 n= 106 n= 161 n= 163
Age > =50 yrs 35.8 32.1 42.2 33.0 40.4 31.3
Army service 63.0 67.0 59.6 62.3 65.8 60.1
Receiving VA disability benefits 83.3 83.5 83.5 83.0 86.3 80.4
VA visit last yr:
Any type 74.1 76.1 80.7 65.1 77.0 71.2
Primary care 63.3 62.4 70.6 56.6 62.1 64.4
Mental health 47.5 50.5 54.1 37.7 51.6 43.6
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 6 of 11
http://www.biomedcentral.com/1471-2288/14/90
items, and there were no interactions between tracking/
privacy assignment and incentive.
As Table 5 shows, Anonymized-Envelope respondents
were substantially more likely than other respondents to
disclose a history of sexual abuse. Several other contrasts
appeared numerically large, even though they did not
reach statistical significance: Anonymous-Postcard respon-
dents reported more childhood physical abuse (p=0.06)
and had fewer positive depression screens compared to the
other tracking/privacy groups (p= 0.09), and Confidential
respondents reported more combat (p=0.09) and had
more positive PTSD screens (p= 0.08).
Table 2 Respondent characteristics by tracking/privacy condition and incentive, results reported as a percentage (%)
Characteristic Overall By tracking/privacy condition By incentive
Confidential Anonymized-Envelope Anonymous-Postcard $10 $20
N= 196 n=61 n=69 n=66 n=85 n= 111
Age > =50 years 44.4 42.6 50.7 39.4 56.5 35.1***
Race
White 52.6 55.7 47.8 54.5 54.1 51.4
Black 27.0 21.3 34.8 24.2 22.4 30.6
Hispanic 6.1 8.2 5.8 4.5 8.2 4.5
Some college experience 74.5 75.4 73.9 74.2 76.5 73.0
Married 67.4 67.8 65.2 69.2 66.3 68.2
Working for pay 61.0 61.0 55.2 67.2 51.9 67.3*
Served in Army 60.2 63.9 56.5 60.6 63.5 57.7
Receiving VA disability benefits:
Per the administrative record 84.2 90.2 82.6 80.3 87.1 82.0
Per self-report 68.4 75.4 71.0 59.1 76.5 62.2*
VA visit in past year:
Any type 78.1 82.0 87.0 65.2* 81.2 75.7
Primary care 67.9 67.2 79.7 56.1 68.2 67.6
Mental health 51.0 54.1 65.2 33.3*** 60.0 44.1
Bold face font signifies a statistically significant difference across group.
*p≤0.05, ***p≤0.001.
Table 3 Characteristics of original sample, stratified by response status and by tracking/privacy condition and
incentive; results reported as a percentage (%)
Characteristic Overall Overall by response By tracking/privacy condition By Incentive
Confidential Anonymized-
Envelope
Anonymous-
Postcard
$10 $20
N= 324 n= 109 n= 109 n= 106 n= 161 n= 163
Respondent? Respondent? Respondent? Respondent? Respondent? Respondent?
Yes No Yes No Yes No Yes No Yes No Yes No
N= 324 n= 196 n= 128 n=61 n=48 n=69 n=40 n=66 n=40 n=85 n=76 n= 111 n=52
Age > =50 yrs 35.8 44.4 22.7*** 42.6 18.8** 50.7 27.5* 39.4 22.5 56.5 22.4*** 35.1 23.1
Army service 63.0 60.2 67.2* 63.9 70.8* 56.5 65.0* 60.6 65.0 63.5 68.4 57.7 65.4
Receiving VA
disability
benefits 83.3 84.2 82.0 90.2 75.0* 82.6 85.0* 80.3 87.5 87.1 85.5 82.0 76.9
VA visit last yr:
Any type 74.1 78.1 68.0 82.0 68.8 87.0 70.0** 65.2 65.0 81.2 72.4 75.7 61.5
Primary care 63.3 67.9 56.2 67.2 56.2 79.7 55.0** 56.1 57.5 68.2 55.3 67.6 57.7
Mental health 47.5 51.0 42.2 54.1 45.8 65.2 35.0** 33.3 45.0 60.0 42.1 44.1 42.3
Bold face font signifies a statistically significant difference between respondents and non-respondents within that column.
*p≤0.05, **p≤0.01, ***p≤0.001.
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 7 of 11
http://www.biomedcentral.com/1471-2288/14/90
Main effects in disclosing sensitive information by in-
centive did not reach statistical significance. However,
there was a trend toward statistical significance in the pro-
portion of respondents randomized to the $10 incentive
with a positive depression screen compared to the $20
respondents (p= 0.08). Among Anonymized-Envelope re-
spondents, those randomized to the $10 incentive were
substantially more likely to screen positive for PTSD than
those in the $20 arm (90.6% v.65.8%;p=0.05). Otherwise,
there were no tracking/privacy-by-incentive interactions.
Table 4 Percentage (%) of respondents fully completing all items in a potentially sensitive scale by tracking/privacy
condition and incentive
Scale/Item Number of
items in scale
Percentage (%) completing all items in the scale
Overall By tracking/privacy condition By incentive
Confidential Anonymized-Envelope Anonymous-Postcard $10 $20
N= 196 n=61 n=69 n=66 n=85 n= 111
Sexual Orientation 1 98.5 100 98.6 95.5 97.6 98.2
Sexual Abuse 8 92.9 93.4 94.2 90.9 91.8 93.7
Other Traumatic events:
Combat
a
26 89.3 86.5 96.8 85.7 89.1 89.5
Childhood physical abuse 5 96.4 96.7 97.1 95.5 95.3 97.3
Past-year events:
Economic hardship 1 98.5 100 97.1 98.5 97.6 99.1
Emotional abuse/neglect 1 99.0 98.4 100 98.5 98.8 99.1
Crime victim 1 99.0 100 98.6 98.5 97.6 100
Physical attack 1 98.5 98.4 98.6 98.5 97.6 99.1
Mental health screens
Depression 5 99.0 100 98.6 98.5 98.8 99.1
PTSD 26 96.9 100 97.1 93.9 94.1 99.1
Problem drinking 5 100 100 100 100 100 100
a
Among those who said they experienced any combat in the Gulf.
Table 5 Percentage (%) of respondents disclosing potentially sensitive information by tracking/privacy condition and
incentive
Sensitive information disclosed Overall By tracking/privacy condition By incentive
Confidential Anonymized-Envelope Anonymous-Postcard $10 $20
N= 196 n=61 n=69 n=66 n=85 n= 111
Not completely heterosexual 7.8 11.5 4.4 7.9 7.2 8.3
Any sexual abuse ever 20.9 14.8 33.3** 13.6 16.5 24.3
Other traumatic experiences
Combat during Gulf War I 78.6 90.2 72.1 74.5 80.7 77.7
Childhood physical abuse 64.3 57.4 59.4 75.8 61.2 66.7
Past-year events:
Economic hardship 43.3 41.0 42.6 46.2 39.8 45.9
Emotional abuse/neglect 22.6 23.3 25.7 18.5 26.2 19.8
Crime victim 10.8 9.8 13.0 9.2 9.6 11.6
Physical attack 5.2 6.7 7.2 1.5 1.2 0.9
Positive mental health screens:
Depression 44.6 50.8 47.8 35.4 52.4 39.1
PTSD 79.2 85.2 77.1 75.8 83.5 75.9
Problem drinking 35.1 31.1 28.6 34.8 36.5 27.7
Bold face font signifies a statistically significant difference across groups.
**p≤0.01.
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 8 of 11
http://www.biomedcentral.com/1471-2288/14/90
Discussion
In this randomized controlled trial, more survey privacy
was not associated with statistically significantly higher re-
sponse rates compared to less privacy, nor did tracking/
privacy condition affect the proportion of respondents
who actually answered our sensitive questions. Instead,
each tracking/privacy condition attracted its own unique
pool of respondents, which in turn may have influenced
our group-specific estimates of sexual abuse, childhood
physical abuse, combat, and mental health problems—
despite the fact that all participants originated from the
same sampling frame. Estimates of sexual abuse, for ex-
ample, were more than 2 times higher in the Anonymized-
Envelope condition than in the other two conditions.
As expected, the higher incentive resulted in a sub-
stantially higher response rate than the lower incentive,
but there was no association between incentive and the
proportion answering our sensitive questions. As with
the tracking/privacy manipulation, each incentive ap-
peared to attract its own unique pool of respondents, with
the larger incentive attracting younger workers for pay
who were less likely to say they were receiving disability
benefits compared to the smaller incentive. Statistically,
prevalence estimates for potentially sensitive or stigmatiz-
ing material did not differ significantly by incentive, des-
pite some numerically large differences. For example,
more than half of respondents randomized to the $10 in-
centive screened positive for depression, compared to
about a third of respondents in the $20 arm.
According to leverage-salience theory [29], individuals
attend to different criteria when deciding to return a sur-
vey and, further, assign to each criterion different weights
and importance. These are known as “leverages”.Inthe
present study, each tracking/privacy and incentive condi-
tion appeared to trigger a different set of leverages, so that
unique subpopulations selectively participated in each of
the study’s arms. When considering sensitive material,
therefore, one cannot assume that the survey method gen-
erating the highest estimate is most accurate.
Since Anonymous-Postcard respondents did not differ
significantly from non-respondents on available measures,
one might be tempted to conclude that this tracking/
privacy method generated the most representative sample of
respondents and hence most accurate prevalence estimates. If
so, one would also have to conclude that the Anonymized-
Envelope approach over recruited sexual abuse survivors. His-
tory of sexual abuse was 13.6% among Anonymous-Postcard
respondents and 33.3% among Anonymized-Envelope re-
spondents. However, we have shown elsewhere that, even
when using Anonymized-Envelopes, survey respondents
underreport their military sexual assault experiences by a
factor of three [30]. This suggests that the Anonymized-
Envelope method either reduces under selection of veterans
with sexual abuse histories or optimizes more “honest”
reporting among those who have such histories—or both—
compared to the Anonymous-Postcard method. It may do
so,however,attheexpenseofeitheroverexcludingvet-
erans with a history of childhood physical abuse or discour-
aging “honest”reporting of childhood abuse. In the present
study the Anonymized-Envelope method generated a sub-
stantially lower, albeit not statistically significant, estimate
of childhood physical abuse of 59.4% compared to the
Anonymous-Postcard’s estimate of 75.8%.
In general, tracking/privacy condition and incentive
level appeared to affect respondent representativeness in-
dependently, with incentives’principal impact being the
recruitment of younger and healthier participants. These
findings may be reassuring to Human Studies oversight
boards, who might otherwise worry that large incentives
coerce the sickest and most vulnerable into survey re-
search participation. Halpern et al. [31] has shown that
higher payment levels do not override research partici-
pants’risk perceptions when considering whether to en-
roll in clinical trials, and, furthermore, poorer, presumably
more vulnerable participants are actually less sensitive to
higher incentive levels than are wealthier participants.
Similar findings have been reported for those deciding
whether to respond to a survey [32].
The present study offers proof-of-concept for pre-merged
questionnaires’utility. However, pre-merged question-
naires will prove most powerful when they incorporate ad-
ministrative information that is highly related to the
survey’s topic (e.g., sexual abuse, childhood abuse) instead
of basic demographic information. Because we did not
have such information for the present study, we cannot
say whether our differing estimates for these sensitive data
across the three tracking/privacy conditions were a func-
tion of reducing or inflating selection biases, a function of
enhancing or impeding “honest”reporting, or both. Future
research will be needed to explore these issues further. It
may well be that different tracking/privacy methods will
prove best for different sensitive topics.
We used a computerized system to manage the track-
ing and administrative data interface in the present
study, but the pre-merged questionnaire concept could
easily be applied to manual methods. For example in a
study using up to three survey mailings per subject, one
could pre-print 3 stickers per subject, file them under
each subject’s name, and then throw away any remaining
stickers once the subject’s postcard or envelope ID was
returned. By study’s end, only non-respondents’stickers
would remain.
Pre-merged questionnaires carry important limitations.
Researchers must be selective in what data they encode
to keep the sticker from becoming uniquely identifying.
If too much information is included, participants might
become identifiable based on their unique combination
of administrative data. We dichotomized age and service
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 9 of 11
http://www.biomedcentral.com/1471-2288/14/90
branch in the present study for this reason. Pre-merged
questionnaires also cannot capitalize on new information.
Health care visits occurring after a survey is mailed cannot
be linked into a dataset, for example. Nonetheless, the tech-
nique offers an advance over usual anonymous methods,
particularly in its ability to assess for non-response bias,
and it could easily be applied to other sensitive topics.
This study’s strengths include its randomized, controlled
design and demonstration of a unique technique to over-
come what has historically been an important limitation of
anonymous methods –namely, an inability to evaluate non-
response bias. We also compared two tracking methods
that can be used in anonymous surveys. Limitations include
its relatively small and unique sample. Since we did not
have access to verifying information, we cannot say how
honestly participants reported their experiences. Findings’
generalizability to other sensitive topics, to non-veterans, or
to women is also uncertain. The study was powered to
examine main effects of incentives on response rates, and
we may have made Type II errors when examining second-
ary outcomes, effects of the different tracking/privacy con-
ditions, and potential interactions. When findings appeared
suggestive, however, we described them in the text. We
also made multiple comparisons, which may have inflated
our Type I error.
Conclusion
We anticipated that greater privacy and larger incentives
would be associated with higher response rate, better par-
ticipant representativeness, more survey completeness, and
greater disclosure of potentially sensitive information. Re-
sults showed no association between privacy and response
rate or survey completeness, supported the association be-
tween greater privacy and participant representativeness,
andyieldedmixedeffectsforthedisclosureofsensitivein-
formation. A larger incentive was associated with higher
response rate and better participant representativeness but
no association with survey completeness. In the intermedi-
ate privacy arm, lower incentive—not higher—was associ-
ated with reporting more PTSD symptoms. Otherwise, we
found no statistically significant associations between in-
centive and disclosing potentially sensitive information.
Having shown that different tracking/privacy conditions
yielded different estimates of sensitive information, we can-
not, unfortunately, tell which estimate was most accurate.
Traditionally, higher disclosure rates of sensitive or stigma-
tizing information have been interpreted as being more
accurate than lower rates, but our data suggest that appar-
ently different disclosure rates may simply be a function of
the subpopulations successfully recruited into a survey.
This possibility needs greater investigation. Pre-merged
questionnaires bypassed many of the limitations historic-
ally associated with anonymous survey methods and could
be used to explore non-response issues in future research.
Abbreviations
ID: Identification number; PTSD: Posttraumatic stress disorder; TWEAK: An
acronym of 5 items used to assess problem drinking: T = tolerance, W =
Worried, E = Eye-opener, A = Amnesia, K = Cut down; VA: Department of
Veterans Affairs.
Competing interests
The authors declare they have no competing interests.
Authors’contributions
MM obtained funding; designed the study; oversaw data collection, analysis,
interpretation; and drafted the manuscript. MAP also assisted in obtaining
funding. MAP, AKB, ABS, SN, and JPG contributed to data collection, analysis,
and interpretation of data. MRP contributed to analysis and interpretation of
data. MAP, AKB, ABS, SN, JPG, MRP read and approved the final manuscript.
Authors’information
MM, MAP, and MRP are core-investigators; AKB is data manager; and SN is
core statistician for the Center for Chronic Disease Outcomes Research at the
Minneapolis VA Medical Center. ABS is a former Center for Chronic Disease
Outcomes Research data manager and currently works in the Health Economics
Program, Minnesota Department of Health, St. Paul, MN. JPG is a former Center
for Chronic Disease Outcomes Research statistician.
Acknowledgements
The Center for Chronic Disease Outcomes Research is a VA Health Services
Research and Development (HSR&D) Service Center of Excellence (Center
grant #HFP 98-001). This study was supported by grant #GWI 04-352 from VA
HSR&D service. The funding agency had no role in the design, data collection,
analysis, data interpretation, manuscript writing, or decision to submit the
manuscript.
Disclaimer
The views presented in this paper are those of the authors and do not
necessarily represent the views of the Department of Veterans Affairs.
Author details
1
Section of General Internal Medicine, Minneapolis VA Medical Center,
Minneapolis, MN, USA.
2
Center for Chronic Disease Outcomes Research,
Minneapolis VA Medical, One Veterans Drive, Minneapolis, MN 55417, USA.
3
Department of Internal Medicine, University of Minnesota School of
Medicine, Minneapolis, MN, USA.
4
Departments of Psychiatry and Psychology,
Minneapolis VA Medical Center, Minneapolis, MN, USA.
5
Department of
Psychiatry, University of Minnesota School of Medicine, Minneapolis, MN,
USA.
Received: 3 March 2014 Accepted: 7 July 2014
Published: 16 July 2014
References
1. Kessler R, McGonagle K, Zhao S, Nelson C, Hughes M, Eshleman S, Wittchen
H, Kendler K: Lifetime and 12-month prevalence of DSM-III-R psychiatric
disorders in the United States: results from the national comorbidity
study. Arch Gen Psychiatry 1994, 51:8–19.
2. Straus M, Gelles R: How violent are American families? Estimates from the
National Family Violence Resurvey and other studies. In Physical violence
in American families Risk factors and adaptations to violence in 8,145 families.
Edited by Strauss M, Gelles R. New Brunswick: Transaction Publishers;
1992:95–112.
3. Resnick H, Kilpatrick D, Dansky B, Saunders B, Best C: Prevalence of civilian
trauma and posttraumatic stress disorder in a representative national
sample of women. J Consult Clin Psychol 1993, 61:984–991.
4. Moum T: Mode of administration and interviewer effects in self-reported
symptoms of anxiety and depression. Soc Indic Res 1998, 45(1–3):279–318.
5. Richman WL, Kiesler S, Weisband S, Drasgow F: A meta-analytic study of
social desirability distortion in computer-administered questionnaires,
traditional questionnaires, and interviews. J Appl Psychol 1999,
84(5):754–775.
6. Rogers SM, Miller HG, Turner CF: Effects of interview mode on bias in
survey measurements of drug use: do respondent characteristics make a
difference? Subst Use Misuse 1998, 33(10):2179–2200.
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 10 of 11
http://www.biomedcentral.com/1471-2288/14/90
7. Testa M, Livingston J, VanZile-Tamsen C: The impact of questionnaire
administration mode on response rate and reporting of consensual
and nonconsensual sexual behavior. Psychol Women Q 2005,
29:345–352.
8. Beebe T, Harrison P, Park E, McRae JJ, Evans J: The effects of data
collection mode and disclosure on adolescent reporting of health
behavior. Soc Sci Computer Rev 2006, 24(4):476–488.
9. Stander V, Olson C, Merrill L: Self-definition as a survivor of childhood sexual
abuse among Navy recruits. J Consult Clin Psychol 2002, 70(2):369–377.
10. Ong A, Weiss D: The impact of anonymity of responses to sensitive
questions. J Appl Soc Psychol 2000, 30(8):1691–1708.
11. Durant L, Carey M, Schroder K: Effects of anonymity, gender, and
erotophilia on the quality of data obtained from self-reports of socially
sensitive behaviors. J Behav Med 2002, 25(5):438–467.
12. Rolnick S, Gross C, Garrard J, Gibson R: A comparison of response rate,
data quality, and cost in the collection of data on sexual history and
personal behaviors. Mail survey approaches and in-person interview.
Am J Epidemiol 1989, 129(5):1052–1061.
13. Campbell M, Waters W: Does anonymity increase response rate in postal
questionnaire surveys about sensitive subjects? A randomised trial.
J Epidemiol Community Health 1990, 44:75–76.
14. Biggar R, Melbye M: Responses to anonymous questionnaires concerning
sexual behavior: a method to examine potential biases. Am J Public
Health 1992, 82(11):1506–1512.
15. Asch D: Use of a coded postcard to maintain anonymity in a highly
sensitive mail survey: cost, response rates, and bias. Epidemiol 1996,
7:550–551.
16. Dunne M, Martin N, Bailey J, Heath A, Bucholz K, Madden P, Statham D:
Participation bias in a sexuality survey: psychological and behavioural
characteristics of responders and non-responders. Int J Epidemiol 1997,
26(4):844–854.
17. Murdoch M, Kressin N, Fortier L, Giuffre P: Evaluating the psychometric
properties of a scale to measure medical students’career-related values.
Acad Med 2001, 76(2):157–165.
18. Murdoch M, Polusny M, Hodges J, O’Brien N: Prevalence of in-service and
post-service sexual assault among combat and noncombat veterans
applying for Department of Veterans Affairs posttraumatic stress
disorder disability benefits. Milit Med 2004, 169(May):392–395.
19. Coxell A, King MB: Male victims of rape and sexual abuse. Sex Marital Ther
1996, 11:297–308.
20. Edwards P, Roberts I, Clarke M, DiGuiseppi C, Pratap S, Wentz R, Kwan I,
Cooper R: Methods to increase response rates to postal questionnaires.
Cochrane Database Methodol Rev 2003, (4). Art. No.: MR000008.
doi:000010.001002/14651858.MR14000008.pub14651852.
21. Murdoch M, McGovern P: Development and validation of the Sexual
Harassment Inventory. Violence Vict 1998, 13(3):203–216.
22. Murdoch M, Pryor J, Polusny M, Gackstetter GD, Cowper-Ripley D: Local
social norms and military sexual stressors: Do senior officers’norms
matter? Milit Med 2009, 174(10):1100–1104.
23. Bernstein D, Fink L: Childhood Trauma Questionnaire: A Retrospective
Self-Report. Manual. San Antonio: The Psychological Corp.; Harcourt Brace
& Co; 1998.
24. Wolfe J, Kimerling R, Brown P, Chrestman K, Levin K: Psychometric review of
the Life Stressor Checklist-Revised. In Measurement of Stress, Trauma, and
Adaptation. Edited by Stamm B. Lutherville, MD: Sidran Press; 1996:198–201.
25. Janes G, Goldberg J, Eisen S, True W: Reliability and validity of a combat
exposure index for Vietnam veterans. J Clin Psychol 1991, 47:80–86.
26. Berwick D, Murphy J, Goldman P, Ware JJ, Barsky A, Weinstein M:
Performance of a five-item mental health screening test. Med Care 1991,
29(2):169–176.
27. Hammarberg M: Penn Inventory for posttraumatic stress disorder:
psychometric properties. Psychol Assess 1992, 4(1):67–76.
28. Russell M, Martier S, Sokol R, Mudar P, Bottoms S, Jacobson S, Jacobson J:
Screening for pregnancy risk-drinking. Alcohol Clin Exp Res 1994,
18(5):1156–1161.
29. Groves R, Singer E, Corning A: Leverage-Saliency Theory of survey
participation: description and an illustration. Public Opin Q 2000,
64(3):299–308.
30. Murdoch M, Polusny M, Street A, Grill J, Baines Simon A, Bangerter A,
Noorbaloochi S, Voller E: Sexual assault of male US Troops during the
Gulf War I Era: prevalence and correlates among applicants for
Department of Veterans Affairs PTSD disability benefits. Milit Med 2014,
179(3):285–293.
31. Halpern S, Karlawish J, Casarett D, Berlin J, Asch D: Empirical assessment of
whether moderate payments are undue or unjust inducements for
participation in clinical trials. Arch Intern Med 2004, 164(7):801–803.
32. Couper M, Singer E: The role of numeracy in informed consent for
surveys. J Empir Res Hum Res Ethics 2009, 4:17–26.
doi:10.1186/1471-2288-14-90
Cite this article as: Murdoch et al.:Impact of different privacy conditions
and incentives on survey response rate, participant representativeness,
and disclosure of sensitive information: a randomized controlled trial.
BMC Medical Research Methodology 2014 14:90.
Submit your next manuscript to BioMed Central
and take full advantage of:
• Convenient online submission
• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution
Submit your manuscript at
www.biomedcentral.com/submit
Murdoch et al. BMC Medical Research Methodology 2014, 14:90 Page 11 of 11
http://www.biomedcentral.com/1471-2288/14/90