Informatics for Pathology-based Specimen Resources. Workshop Summary.
Informatics for Pathology-based Specimen Resources. Workshop Summary.
On May 13 and 14, 1999, the Resources Development Branch held a workshop in Annapolis
Maryland, entitled, "Informatics for Pathology-based Specimen Resources." Participants
included pathologists, informaticians, computer scientists, lawyers, software vendors,
statisticians, and government scientists from NIH, the VA, the Armed Forces Institute of
Pathology, and the CDC. The meeting began with three lectures designed to focus discussion
and provide common concepts and language for the participants: an overview of medical
informatics, by Dr. James Cimino; an update of current activities in pathology informatics, by
Dr. Michael Becich; and a discussion of Distributed Network Query Models, by Joe Futrelle.
Breakout groups addressed the following areas: Uses for pathology informatics systems; Legal
and ethical issues; network technical issues; Political, marketing and economic issues; Model
systems for pathology informatics; and Data and tissue availability. Each breakout group
presented a summary of its discussion to the assembled participants, and time was allotted for
participants to discuss the issues raised by the individual breakout groups.
ASSUMPTIONS AND THEMES
Pathology informatics has served primarily to provide clinicians with access to electronic
pathology reports for their patients. Participants appeared to be in implicit agreement that
additional efforts are needed to expand the role of pathology informatics to make pathology data
and specimens more available to support research. Pathologists were enthusiastic about
becoming involved in this effort and providing pathology resources for this effort. Themes that
emerged repeatedly during the workshop related to how to share specimens and data, how to
standardize data (so that information from multiple sources can be usefully merged), how to
ensure patient confidentiality, and how to provide appropriate incentives and compensation.
STANDARDIZATION OF PATHOLOGY INFORMATICS SYSTEMS
There was agreement on the desirability of institutions using a common or standardized report
format, or at least preparing reports to include all the data necessary for critical text elements to
be automatically extracted and reassembled into a standard report format. Reports should
contain the information indicated by practice guidelines (such as those published by professional
organizations) and data should be entered using common terms. Furthermore, a method (perhaps
automated) is needed to convert the data from standard reports into a standard coding
nomenclature (e.g. SNOMED,CPT, etc.).
The participants cautioned that standardization does not always achieve the desired results. The
Bethesda reporting system for cytopathology was cited as an example. Despite its simple text
approach and wide use, individual institutions have not entered data uniformly or have entered it
inaccurately, producing datasets that cannot be easily used for interlaboratory comparisons.
Other examples of data structure without data rigor are the large multi-institutional systems such
as that used by the VA. While the VA has one of the best structured medical data systems in
existence, with uniform data dictionaries implemented in over 160 institutions, data input on the
local level is sometimes incomplete or inaccurate (e.g. uncorrected misspellings, incomplete or
absent text field entries, etc.).
Despite their reservations, participants enthusiastically endorsed the development of reporting
standards, to provide common entry of text at all institutions. There was further consensus that
most stakeholders in the field of informatics (professional societies, accrediting agencies,
researchers, etc.) would support standardization of pathology reports. As an incentive, third
party payers, including the government, might choose to provide a bonus for pathology reports
adhering to a standard reporting format.
LEGAL AND ETHICAL ISSUES
The breakout sessions on "Legal and ethical issues," agreed that while the legal and ethical issues
related to pathology informatics are beginning to be addressed, many remain unresolved. These
include: specimen ownership; data ownership, and permissible charges for data or tissue access.
Another complicating issue is that there are different requirements for use of specimens and data
by grant supported researchers than for commercial research using the same materials. Many of
the ethical and legal guidelines currently under consideration apply only to federally-funded
researchers. Arrangements between commercial organizations are currently free from certain
restrictive regulations that apply to federal agencies and federally funded researchers.
The group suggested that it would be important to capture consent status, along with any specific
restrictions on that consent, into the electronic pathology record. Informatics initiatives should
tie each and every specimen and data record to current and past conditions placed on use of the
specimen and data. Most importantly, there must be a way to tag specimen reports to indicate
that consent has been withdrawn. This alone is a new informatics activity that should be
considered in developing standardized electronic pathology reports.
The potential usefulness of cancer registries was discussed at length. Forty-nine of the 50 states
have cancer registries, and cancer registries are required by law to collect patient data and to
make it available for research. In the past, pathology data was summarized by the Registrars
during chart review and entered into cancer registry records. Sufficient informatics support may
allow Registrars to automatically acquire complete pathology data linked to specimens. Because
that data would be collected under the legal mandate of the registries, many restrictions that
usually apply to the collection of patient data would be obviated.
There was considerable enthusiasm for an initiative by the NCI to involve the SEER Program
and the CDC in developing new mechanisms to help cancer registries collect information,
including pathology reports as well as information generated outside of the hospital setting, such
as recurrence and treatment data. This would greatly expand the value of both existing cancer
registries and existing pathology data, by improving the clinical information that is collected by
registrars. The important efforts in this area should focus on developing standardized report
formats and the software tools needed to merge cancer registry data with the large pathology
Joe Futrelle opened this discussion with a lecture on the topic of network query models. In
general, network query involves: client software to request information via the internet;
middleware to receive the query and translate it into a standard format with tags to show the
identity or institution of origin, or the security clearance of the client that initiated the query. The
client query, processed by the middleware, is delivered to a target institution. The query is read
and a determination made as to whether or not to answer the query and what, if any, information
to include in the reply. If the query is approved, the institution's own search engine requests the
data from the institutional database and a reply is created and formatted according to a
standardized reply format (such as XML). This process is repeated at each institution that has
agreed to participate. All replies are transmitted to the middleware query agent, with patient
identifiers removed or encrypted. The query agent prepares a report and delivers it to the client.
These steps can be done automatically and virtually instantaneously.
Standards already exist for conducting queries among multiple databases over the internet. Two
of these query models were discussed in some detail; Z39.50 and XML, which is likely to
emerge as the best format for conducting such queries. The workshop participants agreed that
defining software requirements was not as important as defining the data structure. They noted
that developing standard data models is the most important thing that the pathology community
could do. This would require community agreement not only on the names of the data objects,
but on the meanings of those objects. For instance, when "Cause of Death" is selected as the
data object, those entering the data must agree on the meaning of that concept. If a man who
smoked for 30 years develops lung cancer, and dies with metastatic disease, the cause of death
could be "lung cancer," or it could be "long-term cigarette abuse." While the choices are
completely different, both apply and either might legitimately appear as the Cause of Death.
There will be a need for consistent rules for assigning data elements.
MARKETING AND INCENTIVE ISSUES
Pathology departments offering new services, such as informatics and specimen-related support,
should be compensated for providing those services. As an example, one primary reason for the
long and dramatic decline in autopsy rates has been the inability of most pathology departments
to find mechanisms for adequate monetary compensation for autopsies. There was general
agreement that economic considerations have made it increasingly difficult for pathologists to
provide tissue specimens to researchers. In addition, some pathology departments are discarding
archived tissues after 3-5 years of storage. Since pathology departments are reluctant to
undertake new responsibilities without a defined compensation mechanism, they are unlikely to
assume the burden of providing archived tissue with accompanying clinical data reported in a
standardized format without an additional economic incentive.
In addition to specimen handling costs, preparation of reports in a way that supports informatics
initiatives will also require incentives. Third parties might pay a bonus for reports that are
well-coded using CPT, SNOMED or other standard terminologies that permit more accurate
billing, better data analysis, improved outcome studies, etc. Advocacy groups might also be
motivated by the potential to improve research and patient care to provide additional support to
the development of the informatics needed to improve pathology reports. It was also suggested
that laboratory information system vendors might invest in informatics initiatives that appear
likely to add value to their services.
There was enthusiastic agreement that pathology informatics could add significant value to
pathology reports as well as facilitate better research support. Some examples of improved
reporting include: adding images in reports, coding data to make reports more useful to third
parties, using the Bethesda system and other reporting standards in a more consistent manner,
and permanently and securely associating tissue specimens with pathology data.
1. There needs to be a way of associating specimens with clinical data and to do so at
There was general consensus that pathology data has value, that the value of data cannot
be realized until the data is queried, and that data from multiple institutions is needed to
provide the demographic richness and specimen quality needed for translational research
efforts. Any such system must securely protect patient privacy and data confidentiality.
2. NCI needs to convey their research priorities to pathology departments.
There was a strong feeling among pathologists that NCI could play a valuable role in
helping to define those research areas that would benefit from specimen-based
informatics efforts. There was some reluctance to embark on a massive informatics
undertaking unless pathologists knew that their efforts coincided with NCI's interests.
3. There is a need for further discussion of the professional contributions of pathologist and
informaticians to research.
There was concern that collaborative research activities might undervalue the
contribution of those pathologists or informaticians who prepare specimens or data for
researchers. Researchers, institutions, and pathology departments should re-examine the
professional role played by specimen providers and informatics experts. Involvement of
specimen providers and informatics experts in the design of experimental protocols will
be increasingly necessary as specimen-related pathology datasets become available.
4. Standardized pathology reports are needed.
There was complete agreement that NCI should be involved in efforts to define common
data elements and to standardize pathology reports and link the data to the specimens that
needed by cancer researchers. Further, NCI should cooperate with advocacy groups,
professional organizations and laboratory information system vendors to help devise
appropriate incentives for pathology departments to cooperate in report standardization
and the development of specimen-based databases.
5. Cancer Registrars need to include pathology report data in Registry records.
There was considerable enthusiasm for the NCI to cooperate with other governmental
efforts, such as the registry programs of SEER and the CDC, to improve the data that is
routinely collected by cancer registries to support research.
Jules Berman, Ph.D., M.D.
Program Director, Pathology Informatics
Resources Development Branch
Cancer Diagnosis Program, DCTD, NCI, NIH
6130 Executive Blvd
Rockville, MD 20892