Comment Letter on NIH Data
Sharing Proposal
May 10, 2002
Wendy Baldwin, Ph. D.
Deputy Director for Extramural Research
Office of Extramural Research
1 Center Drive, MSC 0152
Building 1, Room 150
Bethesda, MD 20817
Dear Dr. Baldwin:
The Association of American Medical Colleges (AAMC) welcomes
this opportunity to comment on the March 1, 2002 NIH draft
federal policy regarding the sharing of research data. Pursuant
to this policy, applicants for NIH funding would be required
to propose data sharing plans (or offer an explanations for
why data will not be shared) within grant applications. The
adequacy of an investigator's data sharing plan (or the rationale
for declining to share data) would be assessed during the
scientific review of grant applications. The AAMC, while fully
endorsing the laudable aims of this initiative, shares the
concern expressed by the Federation of American Societies
for Experimental Biology (FASEB) that a policy requiring investigators
to devise data-sharing plans when preparing grant applications
is premature and in some cases, unworkable. We also concur
with FASEB that review of an investigator's data sharing plan
should be handled by NIH as an administrative matter external
to the peer review process.
The AAMC represents the nation's 125 accredited medical
schools, nearly 400 major teaching hospitals and health care
systems, more than 105,000 faculty in 98 academic and scientific
societies, and the nation's 66,000 medical students and 97,000
residents. As an association of researchers and research institutions,
we strongly support the principles that underlie the proposed
data sharing policy and endorse efforts by the NIH to facilitate
and encourage the sharing of research data created with federal
funds. Moreover, we believe that establishing standards for
data sharing will become ever more important as biomedical
research addresses increasingly complex problems with technologies
of unprecedented power and scope. At present, however, there
are numerous cultural, technical, and legal obstacles to a
workable uniform federal policy for data-sharing. Certain
of these obstacles are detailed in the comment letter submitted
by FASEB and include the variability of normative data sharing
practices across disciplines, especially in emerging disciplines,
difficulties in defining "final" data at the outset of an
experiment, and the need to protect the intellectual property
rights of investigators and institutions.
AAMC believes that effective policies to promote data sharing
will require creative, discipline-specific solutions to these
complicated problems. We recommend that prior to mandating
data-sharing proposals, NIH should convene advisory panels
composed of experts in the various scientific disciplines
and charge them with devising standards and normative practices
for data sharing within their respective fields of research.
Once consensus is achieved, a grant applicant could be asked
to describe data-sharing plans with reference to agreed-upon
standards for researchers in the applicant's field. In making
this recommendation we note that NIH describes the proposed
data sharing policy as "an extension of the agency's policy
regarding sharing research resources." Those guidelines and
policies were developed by a distinguished ad hoc working
group after lengthy deliberations. The development of proposals
for sharing biomedical research data merits a similarly thoughtful
and deliberative approach.
As the NIH policy statement recognizes, clinical data are
sensitive and require special safeguards to prevent breaches
of patient privacy and confidentiality. Under the new federal
regulatory regime for the protection of medical privacy, the
sharing of clinical data for research may only occur in conformance
with detailed new rules. Entities covered by the privacy rule,
including teaching hospitals and healthcare providers, face
extensive liability for infractions of these rules and must
make difficult judgements about when, and with whom, they
will share patient data for research purposes. Researchers
who create data in a clinical context may not be in a position
to obligate the responsible covered entity, in advance of
the creation of the data, to a data-sharing scheme that involves
unknown recipients and unspecified research purposes.
The proposed NIH data-sharing policy does anticipate that
researchers would remove patient identifiers before sharing
clinical data. Yet the statement fails to acknowledge the
scientific and practical difficulties involved in creating
data sets that are "de-identified" to the exacting standards
of the privacy rule. We are particularly concerned that the
draft NIH data-sharing statement would require researchers
sharing clinical data to "ensure" that individual subjects
of the data cannot be identified. This requirement would be
impossible for any researcher - or any covered entity - to
satisfy. The structure of the "de-identification" provisions
of the medical privacy rule reflects the fact that most patient-level
data are identifiable to an estimable degree of probability
when combined with other publicly available information. The
privacy rule creates avenues for reducing the probability
of re-identification, but does not - and could not - require
covered entities to ensure that subjects cannot be re-identified.
To address the special concerns that arise with the sharing
of clinical data, we suggest that the NIH should fund demonstration
projects to create data enclaves that the Department of Health
and Human Services would certify as "HIPAA-compliant." Such
certification might enable researchers more easily to obtain
needed waivers of authorization under the privacy rule for
disclosure of clinical data to a data enclave. Of course,
any NIH data-sharing policy must recognize that under the
privacy rule, the research use of clinical data disclosed
pursuant to a waiver and stored in a data enclave or repository
would require a protocol-specific waiver or patient authorization.
Thus, mere creation of the enclave or repository would not
ensure that the research community has ready access to the
stored clinical data.
The difficulties inherent in the sharing of clinical data
under the new federal medical privacy rule are emblematic
of the discipline, or field, specific problems that standards
for biomedical research data sharing raise. We would imagine
that equally challenging problems would arise with respect
to developing sharing protocols for nucleic acid and proteomic
arrays, for example. Accordingly, while endorsing NIH's desire
to begin the process of developing such standards, we urge
that the process be undertaken deliberately and in steps,
with adequate input from scientific advisory committees expert
in the particularities of the various research disciplines
and areas. We also remind that efforts to deal with the matter
of biomedical research data sharing began more than a decade
ago under the aegis of the former OSIR. That so little has
been accomplished in the interim perhaps bespeaks the difficulty
of the problem and the need to proceed cautiously.
Sincerely,
Jordan J. Cohen, M.D.
|
