February 2005
Feature Story
Building a timely, more accurate national cancer surveillance
system requires you to think big and then beat the devil in the details
to make it happen.
A pilot project now winding up in Ohio and California and funded by the
Centers for Disease Control and Prevention’s National Program of Cancer
Registries aims to do both. The project’s ultimate vision is a real-time,
standardized way for pathologists to report cancer cases to cancer registries
using a computerized program that incorporates the CAP’s cancer checklists.
The checklists are encoded with concepts from the CAP’s Systematized Nomenclature
of Medicine Clinical Terms, or SNOMED CT, which provides a common language
for capturing, sharing, and aggregating health data.
But the CDC-NPCR pilot, "Reporting of Pathology Protocols for Colon and
Rectum Cancers," is also laying the intricate groundwork for turning that
cyber idea into standard practice. Launched in 2001, the pilot was nearing
completion at CAP TODAY press time and moving into its final phase. A
second project will focus on reporting melanomas as well as breast and
prostate cancers.
Both phases of the CDC-NPCR pilot project make use of the SNOMED-CT-encoded
CAP cancer checklists. The CAP Cancer Committee first published the cancer
protocols (the checklists are the short forms of the protocols) in the
1990s, and have updated them since, to aid surgical pathologists in reporting
malignant tumor specimens completely, accurately, and uniformly. Using
the CAP protocols standardizes the content of reports, and encoding with
SNOMED CT standardizes the meaning.
"If everyone used the format [developed in the pilot] for reporting
and transmitting data, we would be able to process cancer cases more quickly,
and potentially see causes of cancers in real time," says Ken Gerlach,
MPH, CTR, a CDC health scientist and the project officer for the pilot.
Epidemiologists could hone in more quickly on geographical cancer clusters
to search for underlying causes of certain cancers. Some experts say such
a standardized reporting system could have potentially hastened discovery
of the link between human papillomavirus and cervical cancer—or
identified geographical clusters of Kaposi’s sarcoma and lymphoma associated
with HIV infection.
Using a more real-time surveillance system, researchers could more quickly
contact patients with life-threatening cancers to participate in clinical
trials. And the improved accuracy of the reporting system would make it
possible for researchers to compare the exact type of cancers to incidence,
survival statistics, and outcomes to identify best practices in prevention
and treatment.
The CDC-NPCR pilot project’s approach circumvents the
need for cancer registrars to pore over narrative pathology reports to
extract and encode information about a patient’s cancer, as they now often
do to enter cases into their databases. The paper review slows the cancer
data stream and leaves room for errors in translation. "When you use narrative
reports alone, some of the information can be ambiguous," says Bette Smith,
BTM, RHIT, CTR, program manager for the Ohio cancer registry.
What’s more, hospitals and laboratories report cancer using a hodgepodge
of software programs, including homegrown versions, which have varying
data specifications and capabilities.
By contrast, the two anatomic pathology laboratories participating in
the pilot—University Hospitals of Cleveland/Case Western Reserve
University in Ohio and the University of California Irvine Medical Center—report
cases of colon and rectal cancer on a trial basis to the state cancer
registries using computerized synoptic checklists encoded with SNOMED
CT.
"The pathologist toggles through a data entry screen and checks off
the appropriate items to describe a case of colorectal cancer," the CDC’s
Gerlach says. The report also includes a narrative clinical history and
miscellaneous text report, which pathologists can enter themselves or
dictate and have someone else enter, says Joseph Willis, MD, associate
professor of pathology and oncology for the University Hospitals of Cleveland/Case
Western Reserve University.
The software program automatically links each data element in the synoptic
checklist to a SNOMED CT code, a process that’s invisible to the pathologist
completing the cancer checklist. "Pathologists and clinicians don’t go
around talking in SNOMED codes where they say this patient has an MA302
or whatever. They use medical terminology," Dr. Willis notes.
But the SNOMED CT coding ensures the pathologist’s examination of a colorectal
cancer is reported to the state central cancer registry in a standardized
nomenclature. The cases, after they’re consolidated and identifications
are removed, will be transmitted to the CDC-NPCR and the National Cancer
Institute’s Surveillance, Epidemiology, and End-Results, or SEER, program.
"SNOMED goes into depth, so that for every cell type on this checklist
the pathologist reads there is a term universally used and equivalent
to the World Health Organization’s code list," or ICD-O-3, says Barry
Gordon, PhD, director of C/Net Solutions, the maker of cancer registry
software used by the California cancer registries (and the University
of California Irvine in this pilot) and part of the nonprofit Public Health
Institute.
The computerized synoptic format improves the accuracy and completeness
of the colorectal cancer report to the state cancer registry. For one,
Dr. Willis says, the software is a definition-driven process, with the
definitions for the fields embedded in the software. And the pathologist
can’t sign off on the report without completing all of the fields. For
example, "the pathologist cannot omit certain tumor features required
in routine reporting, such as circumferential margins on rectal cancers,
which tell if the surgeon excised the tumor completely," he says.
University Hospitals of Cleveland uses Cerner Corp.’s CoPathPlus system
with Synoptic Reporting to enter and transmit the SNOMED-encoded colorectal
cancer checklists to Rocky Mountain Cancer Data System, a nonprofit cancer
registry software vendor in Salt Lake City serving 22 state cancer registries.
Rocky Mountain translates them into NAACCR format (North American Association
of Central Cancer Registries) before transmitting them to the Ohio registry.
Cancer registries are excited by the work being done
in the pilot. "They see the potential for efficiencies and data accuracy
in terms of bringing in the cases and prepopulating the cancer surveillance
report," Dr. Gordon says. "The data forms some of the core data for reporting
a case." (Cancer registrars would still have to review the medical record
for information about the diagnostic workup and treatment to complete
the report on a patient’s cancer.)
Others see different but equally promising potential. Fritz Lin, MD,
director of surgical pathology and cytopathology at the University of
California at Irvine Medical Center, says the synoptic reporting would
make it possible for pathologists to retrieve and survey cases and conduct
operational research targeting colorectal cancers. "If someone wants to
do molecular genetic studies, for example, they could retrieve the blocks
and have all the information available about the tumor on the synoptic
report," he says.
Yet developing a more real-time, standardized electronic cancer reporting
system is more than a mouse click away, as the pilot study participants
realized at the project’s outset.
"The major challenge involves translating a paper document, developed
initially as a guide for pathologists to report malignant tumors, into
machine-readable codes that can be transferred" to state cancer registries,
says Kathleen Davidson-Allen, a quality control specialist at the California
Cancer Registry.
The CDC-NPCR project brings to the table the various players in pathology,
public health, cancer surveillance, and information technology needed
to figure out how to pull off that feat. Says C/Net’s Dr. Gordon: "Half
of the pilot project has involved defining standards for messages so that
other vendors that want to implement synoptic reporting will have a clear
guide. As people get excited about this type of system, we don’t want
vendors to develop incompatible products, which means we will go back
to being the Tower of Babel."
The most difficult part of the project was combining the expertise of
the CDC, the Ohio and California cancer registries and their software
vendors, the CAP’s Cancer Committee and the CAP’s Division of SNOMED International
(the latter maintains SNOMED CT), and the two anatomic pathology labs
that are transmitting the data for the pilot.
"The group had to look at the way data are coded and how to standardize
information in a way that could be used for computerized functions," says
Georgette Haydu, administrative manager for the Ohio cancer registry.
"Although that required a lot of work and cooperation, it was great to
watch people from different fields unite and come up with a ’programmable’
pathology reporting checklist."
To standardize data submission and transmission, the pilot participants
chose to use Logical Observation Identifiers Names and Codes, or LOINC,
as the codes for the questions or metadata, and the CAP’s SNOMED CT codes
for the answers (or data items) to the questions on the three CAP colon
and rectal cancer checklists. The pathology synoptic reports are thus
encoded with LOINC and SNOMED CT codes and transmitted using Health Level
7, version 2.3.14. HL7 is an information exchange protocol used in health
care and by the CDC’s Public Health Information Network for reporting
other diseases.
Though the pilot study had a couple of months to go at press time, the
Ohio and California cancer registries already had preliminary results
that will help fine-tune the process for phase two of the pilot.
For one, they found they could exchange computerized text messages using
the standardized format, which was a key goal of the project.
The pilot involves a retrospective examination of existing narrative
pathology reports generated by the laboratories involved in the project
to see if pathologists in those labs routinely gather enough information
to complete the CAP colorectal checklists. "The goal," Dr. Gordon says,
"is to look at how much the labs would have to change their operations
to use the synoptic reporting, since this is a study about making pathology
reports more efficient to use by cancer registries." Based on a review
of 50 narrative reports, they found they could code what’s needed for
the synoptic reports— "96 percent of the required data items were
completed," he says.
In performing the retrospective review, the California Cancer Registry
also found a few items in the software program’s colorectal cancer protocol
data set that could not have been captured by looking solely at the narrative,
Dr. Gordon adds. "That means having to complete the synoptic report is
good for encouraging complete compliance with the CAP checklist."
The Ohio cancer registry is having two different cancer registrars abstract
data from the narrative reports at University Hospitals of Cleveland to
compare to reports coming in the synoptic form in the pilot project. "While
the analysis isn’t yet complete, it has thus far shown the synoptic reporting
system to be very accurate," Smith says.
To help anatomic pathology laboratory information system
vendors use the pilot project’s blueprint, the pilot participants are
preparing a report that describes the project, including its method, evaluation,
and conclusions. "The report will also include tables and appendices related
to actual message and transport—or how we packaged the data—so
other people can use it as they develop these kinds of projects," says
CDC’s Gerlach. "Other people in the anatomic pathology and cancer registry
community will be able to access the findings of our pilot project."
The CDC-NPCR has started preliminary work on phase two of the project,
using CAP checklists for breast and prostate cancers and melanoma, which
will employ the same LOINC, SNOMED CT, and HL7 combination for encoding
and transmitting cancer cases to state cancer registries in California,
Maine, and Pennsylvania.
"The second phase of the CDC-funded project will be good because it
covers more reportable cancers at major common anatomical sites," Dr.
Gordon says. "And melanoma is a problematic cancer site because specimens
and diagnoses are handled by so many settings—big and small labs,
physician offices, out-of-state labs, etc." It’s also staged differently
than other cancers, he adds. The CAP’s SNOMED division will be involved
in the project.
Adopting the findings of the pilot project will be voluntary for hospitals
and laboratories reporting cancer. "There won’t be a federal mandate for
people to report cancer using this approach," says Gerlach, "but to be
accredited by the American College of Surgeons’ Commission on Cancer,
a hospital or cancer center has to use the essential data elements of
the CAP cancer protocols." Taking the reporting to the next step using
the synoptic form of the cancer protocols, he says, "isn’t a major leap
if organizations take advantage of the groundwork laid by the pilot projects."
Problems remain in terms of finding a way to report in situ cancers and
benign brain tumors, which cancer registries are required to do. "The
CAP checklists are for invasive carcinomas only," says the California
Cancer Registry’s Davidson-Allen. "And while the CAP breast checklist
does list some in situ histologies, those would have to be in conjunction
with invasive tumors since the checklist applies only to invasive carcinomas."
In that respect, another mechanism would be needed to identify in situ
cancers. "But that’s part of the evolution of the CAP cancer checklists,"
she says.
Dr. Willis of University Hospitals of Cleveland agrees the CAP checklists
are a work in progress. For example, he sees no reason why the checklists
won’t evolve over time to include additional information about tumor characteristics,
such as molecular processes and other prognostic markers. "The checklists
can be expanded in terms of information collected without having to invent
a new system for synoptic reporting and data transmission," he says.
As for predictions about how long it might take to get a real-time national
public health surveillance system in place, the CDC’s Gerlach believes
that could occur "within years" as opposed to decades.
"President Bush is talking about the importance of the electronic health
record as a national priority," Gerlach says. "And as we move in that
direction, we’ll see more electronic reporting in all areas, including
pathology."
Karen Lusky is a writer in Brentwood, Tenn. |
