C R E
D I T : N . H O O G S L A G /I K O N I M A G
E S /A L A M Y
T ere has been an exponential growth in per-sonal health data supplied by urs of mobile devices, health apps, and the social Web (so-cial networking sites, online dia support groups, and health-related information sites). T e sources and data from tracking of con-sumers’ online behavior coupled with ad-vanced bioinformatics tools of er opportuni-ties for u in health rearch [for examples, e (1, 2)]. Tweets about dia outbreaks have been correlated with of cial public health surveillance data and have proven to be an important source of early outbreak de-tection (3).
A central ethical question is whether in-dividuals who have provided personal in-formation online in nonrearch contexts have connted to rearch us. Here, we explore the issue of informed connt for health rearch performed using information collected from the Web, discuss some limita-tions of current practices, and of er recom-mendations for improving connt practices through a more tailored, context-nsitive approach that makes u of the dynamism of the Web-bad context. Our proposals are rooted in the ethical imperative of protecting individual rights and respecting autonomy w
hile enabling a dynamic rearch environ-ment for the advancement of clinical medi-cine and public health.
A CHANGING CONTEXT FOR INFORMED CONSENT
Health-related rearch proposals with hu-mans typically undergo prospective review by rearch ethics committees that ensure that the study is designed and conducted in an ethical manner that protects the privacy and autonomy of individuals through the informed connt process, balances risks
and benef ts, and ensures that subject lec-tion is equitable. Informed connt requires that potential participants are provided with adequate information to make an informed and voluntary decision about rearch par-ticipation. Despite the standard practices of obtaining connt, there is a prevalent notion that the process is broken (4). How can an ad-equate connt process be achieved in health rearch that involves data collected in the Web environment?
白贝怎么吐沙Currently there is limited ethics guidance specif cally for rearch with data collected on the Web (www.aoir/reports/ethics2.pdf). Ref ected in the dearth of ethics guide-lines is either a lack of acknowledgment of this growing area of rearch or perhaps a n that this rearch should not b
e treated dif erently from other conventional areas of rearch. Further, although Web-bad rearch is inher-ently interdisciplinary becau
of the range of rearch areas and diver sources of data, limited dialogue exists among the rep-rented disciplines in terms of a common t of rearch princi-ples. New sources of online data and innovative health rearch applications and the increasingly disparate sources of data chal-lenge traditional approaches to informed connt.
Conventional informed con-nt models are ill suited be-cau they were not conceived in the context of the evolving applications and functionalities of social media that enable in-novative rearch designs. In addition, traditional approaches to rearch ethics and informed connt include an ethical dis-tinction between public and private information: T e u of
publicly available information typically does not require in-formed connt of the individual, whereas
the u of “private” information may require connt depending on whether the informa-tion allows an individual to be identif ed. In an online world, the public-private distinc-tion is increasingly blurred. Should explicit connt be required if a rearcher collects and analyzes deidentif ed Facebook posts that reveal health status or health behaviors? Can health status information shared on social networking sites for patient commu-nities (for example, ) be ud for rearch without individual in-formed connt? Is such information prop-erly characterized as public or private?
Current discussions about the public-private dichotomy in the online world include a newer, richer concept that views privacy within “contextual integrity” (5). T is ap-proach argues for understanding the impor-tance of the context in which information is located, and determinations of acceptable u are informed by expectations for the u of information within the context in question. T is approach places heavier emphasis on the intent of individuals regarding access to per-sonal information rather than on the tradi-tional approaches that demand that rearch-ers protect privacy as a condition of rearch.Equally challenging to traditional con-cepts of informed connt is the control of
P O L I C Y
Caught in the Web: Informed Connt for Online Health Rearch
Ef y Vayena,1* Anna Mastroianni,2 Jef rey Kahn 3
*Corresponding author. E-mail: vayena@ethik.uzh.ch
1Institute of Biomedical Ethics, University of Zurich, Zurich 8032, Switzerland. 2School of Law and Institute for Public Health Genetics, University of Washington, Seattle, WA 98195, USA. 3Berman Institute of Bioethics, Johns Hopkins University, Baltimore, MD 21205, USA. A context-specif c appr
oach to informed connt for Web-bad health rearch can facilitate a dynamic rearch enterpri and maintain the public trust.
Mission control. Data sharing for health rearch requires an environment of trust.
o n F e b r u a r y 21, 2013
s t m .s c i e n c e m a g .o r g D o w n l o a d e d f r o m
personal information on the Web. When sub-jects are traditionally asked to connt to the rearch u of their data, the limits of that u are spelled out in detail. Personal data posted on or collected by Web sites, how-ever, can be sold or shared and subquently ud in rearch; thus, it is nearly impossible for urs to maintain control of their data, its dif usion, and subquent us. As such, the notion of connting to rearch u of data los meaning when the u can involve many unknown rearchers and us in per-petuity. Such open-ended u of data renders the well-established right to withdraw con-nt to collection and u of personal data for rearch meaningless. A recent controversial European Union proposal attempts to ad-dress the “genie out of the bottle” problem by revising data privacy standards to include a digital “right to be forgotten”—urs must be granted an option to delete personal data from the Web permanently (ec.europa.
eu/justice/data-protection/document/re-view2012/com_2012_11_en.pdf).
DO WEB USERS CONSENT?
Health rearch using the Web is gaining momentum regardless of available guid-ance. How informed connt is treated and what it means varies substantially by site and project.
T ere are two broad types of data gath-ered through the Web that can be ud in health-related rearch, either by the Web site owner or by third parties: (i) informa-tion actively supplied by the ur (medical histories, genomic data, and Web posts), and (ii) personal information collected by the Web site while the ur is interact-ing with the site (IP and e-mail address, arches, and location data). Both data types may be required by a Web site to enable it to provide the promid rvices to the ur. Many Web rvices are provided to the ur free of charge, while the content that urs generate and personal data they provide be-come trading capital for the companies that provide the rvices. Web sites of en autho-rize third parties to access their data ts for commercial and rearch purpos.
T e disclosure to urs of the potential us of personal data vary dramatically from site to site. Further, no publicly available studies have yet documented whether urs understand or are even a
ware of the poten-tial us of their data when they access a site. In reviewing a range of Web sites (6) that collect or contain health data for rearch, we have identif ed three general approaches to connt: (i) rearch participation as a
condition of u of the site, (ii) opt-in to re-
arch, and (iii) opt-out of rearch.
In what we term “condition of u” re-
arch participation, Web sites state in their
terms of u, terms of rvice, or privacy
statements that they maintain the right to
u the data they collect for rearch, among
other us. By virtue of using the site, the
ur agrees to rearch participation. T is is
壁纸唯美equivalent to a so-called browwrap agree-
ment, whereby the ur agrees to the terms of
u without any af rmative conduct, such as
clicking an “I agree” button (www.ef /wp/
clicks-bind-ways-urs-agree-online-terms-
rvice). It is most likely a carryover from the
consumer-oriented sites that u browwrap
approaches as a basic disclosure of policy but
without real expectation of careful review
and af rmative connt.
T e condition of u approach rais three
potential concerns. First, the ur provides a
general connt to a range of us, including
rearch, rather than connt for a specif c re-
arch project or rearch u, and is unable
to access the site without giving broad gen-
eral connt. Second, possible rearch u
is of en (but not always) listed among many
other us within the boilerplate language
of disclosures and indemnif cations, mak-
ing it questionable that the reader will take
notice. T ird, the condition of u approach
was craf ed for consumer agreements to Web
site u rather than to accommodate the re-
quirements for informed connt in rearch.
T us, the condition of u approach stands in
stark contrast to the conventional approach
to informed connt in health rearch: a
process carefully constructed so that (i) in-
dividuals are adequately informed about the
project, (ii) the meaning of rearch partici-
pation is clear, (iii) the potential participant
makes a voluntary agreement to participate,
and (iv) the potential participant is of ered
the option to withdraw from the rearch.
On this analysis, condition of u rearch
participation does not meet the standards of
informed connt except in the most limited
and legalistic understanding of connt as
agreement evidenced by accepted terms of
u.
By contrast, the opt-in approach enables
urs to agree to participate in a specif c re-
arch project. Web sites that u opt-in may
include a statement with information about
the project followed by a link that leads to the
project or a requirement to click an “I agree”
button to allow rearch u of personal data.
In contrast to condition of u, opt-in par-
ticipation requires an af rmative decision
by the ur before participating in rearch.
T is is equivalent to the common approach
to sof ware or other licend goods in the
online world, in which a ur must agree
to a licensing agreement (also known as a
clickwrap agreement) before accessing the
product or site. Y et, in that consumer con-
text, rearch shows that urs spend almost
no time browsing the text of the agreement
before clicking the box, making it unlikely
that urs opting in to rearch using this
model will carefully read the agreement texts
(7). T us, allowing a rearch participant the
opportunity to review information about
the potential rearch u and conf rming
participation with an af rmative act comes
clor to satisfying the conventional criteria
of informed connt, but it is far from clear
that the opt-in model achieves the goal of in-
formed voluntary rearch participation.
With the opt-out approach, Web site us-
ers agree to rearch us of their data un-
less they take action to exclude themlves
from participation. Urs thereby control
their data, provided they are aware that the
opt-out option exists. For example, some
arch engines of er urs the option of opt-
ing out of tracking so that personal data are
not collected or stored, although urs typi-
cally must be suf ciently web savvy to locate
the opt-out option. T e opt-out approach
to informed connt exists in conventional
biomedical rearch when the study pos母亲的伟大
low risks or when obtaining opt-in connt
is impractical and could undermine study
design (for example, large-scale epidemio-
logical studies and genome-wide associa-
tion studies). In the limited contexts, the
u of opt-out approaches has been thor-
oughly debated in the literature and by re-
arch ethics committees before being put
年年岁岁朝朝into practice. Rearchers in traditional
ttings have advocated for wider accept-
ability of this opt-out approach to informed
connt, arguing that it facilitates rearch
while still safeguarding autonomy (www.
ieaweb/index.php?option=com_cont
ent&view=category&id=22&Itemid=54).
Whether an opt-out approach to connt in
health-related Web-bad rearch satisf es
conventional criteria of informed connt re-
quires analysis by consumers, scientists, legal
scholars, and rearch ethics committees.
RECOMMENDATIONS
T e health rearch enterpri continues to
evolve along with new possibilities for har-
nessing online personal health data from
o
n
F
e
b
r
u
a
r
y
2
1
,
2
1
3
s
t
m
.
s
c
i
e
n
c
e
m
a
g
.
o
r
g
D
o
w
n
l
o
a
d
e
d
f
r
o
m
the social Web. Y et, such rearch is still gov-erned by rules t to address the issues faced in traditional clinical rearch; this creates a mismatch between the policy requirements for rearch protections and the types of is-sues faced in health rearch using online data. It is important and timely to develop guidance that specif cally address this new rearch context. T e values that underpin conventional health-rearch rules remain important. Respect for individual autonomy, balanced and equitable distribution of risks and benef ts, and the right to privacy have not lost normative weight becau the social Web challenges the approaches we have cre-ated to respect them. On the contrary, the challenge prents an opportunity to explore more nuanced and innovative ways of inter-preting the values.
It is hard to imagine a uful “one size f ts all” connt model for rearch in the Web environment. Appropriate connt models will depend on the mission of the site, nsi-tivity and identif ability of the data collected, purpo of the rearch, and risks and ben-ef ts of participation. An interactive process is better suited to meeting the criteria of in-formed connt. At a minimum, transparent disclosure of the rearch us of online per-sonal data are required.
Two types of connt models are well suited for health-related rearch on the Web. T e f rst is bad on the relatively new concept of the Portable Legal Connt (PLC), a legal framework for rearch con-nt developed by the Connt to Rearch project (weconnt.us; www.inspire-2live; www.sageba). It allows participants who are willing to relinquish control of their personal information to at-tach a one-time rearch connt to their health and genetic data, which they upload themlves onto the Web site. T e data can then be ud for rearch purpos by any rearcher who agrees to specif c criteria: publication of rearch results in an open-access forum, no reidentif cation of partici-pants, and no redistribution of data unless the data recipient agrees to the PLC condi-tions. T e overarching goal of the PLC is to minimize barriers to data sharing and make rearch data more widely available. Trans-action costs related to contacting partici-pants for connt to individual studies are eliminated, privacy concerns are minimized as the data are deidentif ed to the extent possible, and the participants are informed of potential privacy risks before connting.
Participants may withdraw their data from
the databa at any time, but are clearly ad-
vid that once data are uploaded, it may not
be possible to remove it from all sources (for
example, from rearchers who have already
downloaded, shared, or ud the data). T e
Self Contributed Cohort for Common Ge-
nomics Rearch Study (SCC-CGR) (
weconnt.us/about-us) is the f rst to imple-
ment the PLC.
T e PLC improves on existing connt
process in two important ways: (i) the con-
ditions, rules, and restrictions that apply to
all rearch and to rearchers who access
the data are transparent, and (ii) the risks to
participants of sharing their data are clearly
articulated. However, in contrast to tradi-
tional connt models, the PLC approach, in
its propod state, has some apparent short-
comings: (i) PLC participants must be willing
to give up control over their personal health
data, including a limited right to withdraw
and the choice to decline specif c rearch
projects; (ii) deidentifying of data limits its
ufulness for some rearch projects; (iii)
PLC relies on a lf-lected, well-informed
population of computer-savvy urs who are
unlikely to be reprentative of the popula-
tion at large; and (iv) PLC cannot be ud for
Web data collected for nonrearch purpos.
T e shortcomings of the traditional con-
nt process and PLC argue for an approach
that is nsitive to the unique aspects of Web-
暖心人
bad health rearch and that harness the
dynamic aspects of the Web environment.
Collaborative and context-specif c connt
employ the communicative and real-time
features of the Web to facilitate a more dy-
namic approach to informed connt (8).
Instead of the traditional approach of a one-
time agreement that includes boilerplate text,
a ur could receive tailored information on
rearch participation with specif c choices of
options relevant to his or her situation. Col-
laborative connt might provide a way to ad-
dress a problem that has challenged even the
traditional connt process: how to design
ef ective ways of communicating informa-
tion to prospective participants. Moreover,
transparency requires commitment to clarity
and the provision of accurate and appropri-
ate information by rearchers. Sites engaged
in health rearch or that allow third parties
to u their data for rearch must modernize
communication by making u of the multi-
media capabilities af orded by the Web.
PLC will give individuals a way to share
(or cede control of) their health-related
data. But what about prospective partici-
pants who desire greater control? Empirical
evidence shows that people care about the
way their data are ud (9). Giving the ur
control will also contribute toward building
trustworthy relationships and likely increa
ur participation in rearch. T e “health-
information altruist” (10) is willing to con-
tribute to rearch for the common good.
T e claim behind the health data rights
movement is that individuals should be able
to control the sharing of their information
(www.HealthDataRights). T e trends
can result in valuable contributions to re-
arch only if encouraged by an environment
that is conducive to trust.
REFERENCES AND NOTES
1. B. H. Wagenaar, P. S. Sullivan, R. Stephenson, HIV knowl-
edge and associated factors among internet-using men
who have x with men (MSM) in South Africa and the
United States. PLoS ONE7, e32915 (2012).
2. J. Frost, S. Okun, T. Vaughan, J. Heywood, P. Wicks, Pa-
tient-reported outcomes as a source of evidence in off -
label prescribing: Analysis of data from PatientsLikeMe.
J. Med. Internet Res.13, e6 (2011).
3. R. Chunara, J. R. Andrews, J. S. Brownstein, Social and
news media enable estimation of epidemiological pat-
terns early in the 2010 Haitian cholera outbreak. Am. J.
Trop. Med. Hyg.86, 39–45 (2012).
4. E. C. Hayden, Informed connt: A broken contract. Na-
ture486, 312–314 (2012).
5. H. Nisnbaum, A contextual approach to privacy online.
Daedalus140, 32–48 (2011).
6. E. Vayena, A. Mastroianni, J. Kahn, Ethical issues in health
rearch with novel online sources. Am. J. Public Health
102, 2225–2230 (2012).
7. R. Boehme, S. Köpll, Trained to accept? A fi eld experi-
ment on connt dialogs. In Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems
(ACM, New York, 2010) pp. 2403-2406; dl.acm/
citation.cfm?doid=1753326.1753689
8. P. Bernal, Collaborative connt: Harnessing the
strengths of the internet for connt in the online en-
vironment. Int. Rev. Law Comput. Technol.24, 287–297
密码登录
(2010).
9. J. Turow, J. King, C. J. Hoofnagle, A. Bleakley, M. Hen-
nessy, Americans reject tailored advertising and three
activities that enable it (University of Pennsylvania, An-
nenberg, 2009); /sol3/papers.
cfm?abstract_id=1478214
10. I. S. Kohane, R. B. Altman, Health-information altruists—
A potentially critical resource. N. Engl. J. Med.353, 2074–
2077 (2005).
Competing interests: The authors declare that they have no
competing interests.
Citation: E. Vayena, A. Mastroianni, J. Kahn, Caught in the Web:
Informed connt for online health rearch. Sci. Transl. Med.5,
173fs6 (2013).
10.1126/scitranslmed.3004798
o
n
F
e
b
r
u
a
r
y
2
1
,
2
1
3
s
t
m
.
s
c
i
e
n
c
e
m
a
g
.
o
蚕的生长过程图片r
g
D
o
w
n
l
o
imap服务器a
d
e
d
f
r
o
m
