Never underestimate what a little competition and awareness can do.
Data published last year in the Journal of the National Cancer Institute indicated that only 44 percent of those participating in the NCI’s Surveillance, Epidemiology, and End Results program who were diagnosed with colorectal adenocarcinoma in 2001 received adequate lymph node evaluation (J
Natl Cancer Inst. 2005;97:219–225). Though this number was higher than that in 1988, when only 32 percent of colorectal cancer patients had the suggested number of lymph nodes retrieved, the researchers observed that the number was still low, leading them to conclude that local surgical and pathology practice patterns may have an effect on whether lymph nodes are adequately evaluated in patients with colorectal cancer.
These data confirmed what John Sorge, MD, a pathologist in the Department of Pathology at Rex Hospital, Raleigh, NC, and his colleagues from Pathology Service Associates had suspected when they began in 2003 to compare the performance of two of PSA’s pathology practices in lymph node recovery from colorectal specimens.
“Pathology Service Associates is recognized as a provider of business services and support to independent pathology practices, but one of our purposes in performing this project was to demonstrate that we are also committed to improving the science of pathology,” Dr. Sorge says.
Dr. Sorge was part of a group of pathologists from PSA who met in early 2002 to discuss what could be done to improve quality in their practices. In the group was E. Eugene Baillie, MD, a pathologist at Piedmont Pathology Associates and Anderson (SC) Area Medical Center and also president of InReach Corp. Dr. Baillie volunteered the use of InReach’s software to coalesce and compare data from the two organizations’ computer systems in an effort to provide benchmarking, quality assurance, and process improvement opportunities.
“We knew we had a number of disparate methods of reporting among the various practices in Pathology Service Associates, and we believed our software technique was something that could potentially be applied not only to other practices, but to other disease states,” Dr. Baillie says.
At first, the group developed a large list of data points to compare. With Rex using the CoPath Plus system and Anderson using Cerner Millenium, the institutions were working from different templates, computer systems, and report arrangements. Thus, as they downloaded their data into the single database for data mining, the data had to be standardized. Because the InReach software uses extensible markup language, or XML, each site was able to mark the important fields and define the corresponding text in the underlying database. This is called source labeling because only the information taken directly from the source is marked.
Once the input from the data sources was standardized, a pre-processing step was performed and fields were added based on the information contained in the source fields. “Between the two of us,” Dr. Baillie says, “we had the same data elements but not necessarily in the same part of the report or in the same order.” After combining their databases, the uniformly labeled data points could be compared from one institution to the next.
With the system, the two organizations can now view all data from their own site but only non-sensitive data from the other site. “The software has a feature that allows us to read our own data, but we can’t read anybody else’s data,” Dr. Baillie says. “We hide all of the patient or institution information so that we are HIPAA-compliant.” The software also makes it possible to construct studies that calculate aggregate statistics across both sites, and allows for all existing data to be reevaluated if different or new information is added.
Since all of every pathology report for both institutions was in the database, the InReach software and computer generated so much data on selected line items related to even breast, prostate, colorectal, and lung cancer that it was impossible to evaluate all of it. That’s when the researchers decided to narrow the scope of their project. “We had to determine what we felt would lead to process improvement and have the greatest impact on quality in our practices,” Dr. Sorge says. “We chose colorectal carcinoma because I had been to several presentations at which speakers had indicated that consulting pathologists were seeing large numbers of consult cases in which the node numbers for colorectal cancer specimens were not sufficient.”
A CAP Working Group published in 2000 a guideline recommending that at least
12 regional lymph nodes be obtained per specimen from a radical colorectal
resection (Arch Pathol Lab Med. 2000;124:958–965). The American Joint Commission
on Cancer’s Cancer Staging Manual, published in 2002, recommends retrieving
seven to 14 lymph nodes per specimen, though it permits retrieval of fewer
lymph nodes for patients treated with preoperative radiation.
Given these guidelines, the two pathology groups launched a two-year project to compare the number of lymph nodes per specimen they were collecting and to compare their data. Pathology Service Associates, Florence, SC, provided $30,000 to help fund the project, and a Rex Cancer Grant from Rex Hospital supplied $45,000. InReach Corp. donated the software.
The two groups realized early in the planning stage that not only would they be comparing data from disparate information systems, they also would be comparing data collected using two different methods for sampling lymph nodes. The literature indicates that xylene clearing produces the greatest harvest of lymph nodes from a specimen, Dr. Sorge says, but pathology practices are moving away from using it. “Xylene is a solvent that we have tried to remove from the laboratory for years because of its carcinogenic properties,” he notes. His pathology group began three years ago to treat colon specimens with a clearing agent called Dissect Aid, a fixative that turns lymph nodes white in the pericolic adipose tissue, potentially making it easier to find them.
“Anderson was still using the manual method, which gave us the opportunity to not only compare results between institutions, but between entities using Dissect Aid versus the manual method,” Dr. Sorge says.
Another difference noted early on, Dr. Baillie says, was how pathologists state whether or not nodes are involved and the numbers of nodes, so the software had to intelligently “read” the appropriate report sections.
When the first data were extracted from the InReach system at the end of 2003,
they showed that Rex was recovering more lymph nodes per specimen than Anderson,
with a mean number of 12.9 lymph nodes per specimen using Dissect Aid compared
with a mean of 8.8 lymph nodes per specimen at Anderson using the manual method
(Am J Clin Pathol. 2005;124:83–88).
“At the beginning of 2004, we decided to look at what would happen if Anderson began to treat colorectal specimens the same way Rex does,” Dr. Baillie says. Anderson therefore began using a fat clearance product similar to Dissect Aid to help it identify more lymph nodes, and changed the policy governing the harvesting of lymph nodes.
“Prior to the study, we had a policy that left it up to the discretion of the pathologist to go back and retrieve more lymph nodes if they didn’t think they initially harvested enough,” Dr. Baillie explains. “When we began using a fat clearance method similar to that used by Rex, though, we changed the policy so that everyone would cut the specimen initially, but not sample any lymph nodes until the fixation aid allowed easier identification of nodes.
Anderson now has a specific way to cut the specimen and put it in the solution. “Instead of cutting specimens one day and getting reports out the next, we are cutting the specimen, getting the initial sections of the cancer, and then waiting until the next day to have the lymph node count,” Dr. Baillie says. It has delayed the reports by one day, but the whole case is ready at the time the pathologist is reviewing the entire case, he says. Though there will always be cases with fewer lymph nodes, if fewer than 11 lymph nodes are collected using the new process, Anderson must now go back to the specimen to see if there are additional lymph nodes after the specimen has been sitting in the fat clearance solution for a day.
Dr. Baillie says InReach Corp. is ready to launch the software application for those who want it. “We set out with the idea that we wanted to have a system that could evaluate new parameters or new situations or review existing data and reevaluate it—and that’s what we have.”
Ever since the two began comparing data and Anderson changed its process, the number of lymph nodes retrieved per specimen in both pathology practices has been rising.
“To have our numbers compared with those of another organization is very helpful,” Dr. Sorge says. “If we hadn’t done this, I would have no idea how we compared at Rex with anyplace else in the country.”
Dr. Sorge cautions that the study’s take-home message is not that using fat-clearing solutions is better than using manual methods, but that measuring the number of lymph nodes your pathology practice is retrieving and comparing the data to those collected by peers provides an opportunity for benchmarking and process improvement. “I’m certain there are other fat-clearing solutions that will help with lymph node recovery. I’m also certain there are groups who like to use the manual method,” he says. He doesn’t find fault with the latter, of course, so long as those who use it post adequate numbers of lymph nodes retrieved. “It’s all about accurately staging cancer to provide the best patient care,” he adds.
When Dr. Sorge examined the most recent numbers, he discovered that the average number of lymph nodes per specimen he and his colleagues at Rex collected jumped from 16.3 in 2004 to 19.5 in 2005. Anderson’s numbers are up as well, Dr. Baillie says.
Says Dr. Sorge, “We’ve also moved more specimens out of the category in which the number of lymph nodes identified per specimen is less than 11.” He chose that threshold as a compromise between the CAP and AJCC recommendations, though the number is closer to the CAP recommendations. “In 2003, we found less than 11 lymph nodes in 45 percent of our cases. We’ve cut that number down to 10 percent in 2005.”
Reducing that number below 10 percent will pose a new challenge, Dr. Sorge says, because of nonpathology-related barriers, such as specimen length and other patient factors. The computer data showed that, generally, the shorter the length of the specimen, the lower the number of lymph nodes that will be recoverable.
Still, Dr. Sorge remains upbeat about the project because its two hypotheses proved correct. First, that process improvement will move toward the institution with better lymph node recovery. And, second, that lymph node recovery will improve across both institutions with a healthy dose of competition and greater overall awareness.
Sue Parham is a writer in Edgewater, Md.