April 2012
Feature Story

Anne Paxton

Ayottabyte, a unit of data storage containing one septillion bytes, represents a quantity so high that, as of the writing of this article, the next higher magnitude unit had not been named. By some estimates, the current sum of human knowledge might fit snugly inside a yottabyte. Yet the Pentagon’s worldwide communications network, known as the Global Information Grid, already expects to have to manage yottabytes of data by 2015. It’s one example of how advances in computing technology, stunning as they may seem, can easily find practical applications in the field—and soon be stretched to their limits.

IT has been no slouch in the laboratory arena, but anatomic pathology is proving it has data management needs that are constantly pushing the envelope. On the mundane level of volume alone, says Ulysses Balis, MD, associate professor and director of the Division of Pathology Informatics at the University of Michigan Health System, “we’re definitely growing as a specialty.” The guidelines for screening for colorectal and prostate cancers, for example, have increased the number of biopsies as well as cores within cases.

At the same time, ancillary and primary tests have become more complex. “We’re doing significant numbers of immunostains, molecular studies, and gene rearrangements, and for renal studies we have the multi-structural studies. And the complexity will continue to go up as there is more and more literature coupling the promise of personalized medicine with better predictive outcomes. So the AP laboratory information system has to be sufficiently rich and complex to house that kind of data,” Dr. Balis says.

In the quest to keep pace with pathologists’ IT needs, some IT companies are finding that partnerships are a big plus. Aperio is one example. “Supporting access to the evolving world of electronic pathology through the AP LIS is important and achieved through partnerships,” says Mark Wrenn, Aperio director of professional services. “The LIS is the portal into AP, and we’ve done a number of interfaces with partners Cerner CoPathPlus, Sunquest CoPathPlus, what’s now called Sunquest PowerPath, and McKesson Horizon AP.” Aperio announced last fall it would collaborate with Psyche Systems for end-to-end LIS integration with scanning technology for workflow support.

There are two kinds of interactions between laboratory information systems and what Aperio calls its ePathology solutions, Wrenn explains. “The first is the exchange of data; the second is workflow integration. Years ago, we started putting indicators in the LIS to indicate that an ‘eslide’ is available. The user could then select icons to open our viewer. The key interaction there was the LIS to the slide viewer. This year, with Aperio’s eSlide Manager software [release 12], we are moving to more advanced pathology workflows, such as the ability to do intraoperative consults, secondary consults, or tumor boards.” All are possible with this software, he says. “So the interaction is no longer between the LIS and the eSlide viewer, but is now between the LIS and pathology-specific workflows.”

What Aperio is planning in partnership with Psyche Systems, via Psyche’s WindoPath AP system, is an interaction whereby “I will be able to click a button and say, ‘I want to perform an analysis or intraoperative consultation or secondary consult.’ Then the user interface transfers to our ePathology software and that leads the pathologist through a workflow.”

Aperio has created two major solutions, Wrenn says: network and precision—network providing the ability to share images both inside and outside an organization, and precision providing the ability to use image analysis to get better results. Supporting these solutions are services. “We’ve discovered that ePathology is so new that providing services to guide customers in the use of ePathology solutions will be important.”

During presentations, when Wrenn shows a photograph of a pathology desk surrounded by teetering stacks of paper on the floor, the pathologists in the audience laugh. “Each stack represents a case, and this is the state of affairs for many pathologists, and it’s obviously fraught with errors,” Wrenn says. “So having an ePathology system integrated with the AP LIS is important in reducing the opportunity for error.”

What distinguishes Psyche Systems from some other LIS vendors, says CEO Lisa-Jean Clifford, is that “Our product WindoPath focuses on the AP laboratory but shares the database with our clinical LIS, so our customers are able to perform all of their management reporting without having to piece together reports from disparate systems.” While the clinical laboratory has requirements and workflows that are broader than those in anatomic pathology, she says, the AP lab is far more complex than the clinical lab. “So we decided to build the clinical functionality within our AP system.” WindoPath has the ability to incorporate fully digitized slides. “We’ve integrated seamlessly with Aperio to be able to share data and to do single sign-on between both of our systems.”

The AP information system has evolved from being largely an afterthought for most hospitals five years ago, Clifford says. “AP physicians were basically given a system geared toward a hospital facility or clinical LIS and told to make do. Pathologists either had to rework their workflow to accommodate that system, or had to supplement with manual tasks to make the system work for them.” Now the AP lab is being recognized as the complex entity it is, so it is getting the workflow and tools to accommodate the methods and test menus in a technologically advanced way. That digital pathology companies like Aperio are looking to partner with AP LIS companies reinforces Clifford’s belief that digital pathology systems could never replace an AP LIS. “Pathologists who are moving toward digital slide imaging need a system that can incorporate, handle, store, manage, and annotate those images.”

The role of companies like Psyche is to help the AP laboratory as it becomes more complex, integrated, and automated, Clifford says. “The AP lab is really becoming a key in not only diagnosing patients but also helping to manage and monitor outcomes for patient care and treatments. And a truly automated lab will need to capture and store that data as well as have it stored in a manner where it is able to be mined and analyzed in a way that it can help predict future treatment plans that will have better patient outcomes.”

A much tighter integration of the LIS into bench workflow will be a key feature of the next-generation AP system that the University of Michigan Health System is developing in partnership with SCC Soft Computer, says the university’s Dr. Balis. Using the Cerner Millennium PathNet, and what the partners are calling a “wraparound database” for molecular, which is in the Oracle Database 11g release, the systems interoperate in real time for custom interfaces.

But when a systemwide upgrade takes place at the university, as scheduled, Dr. Balis expects even further advances. Under the traditional model, the LIS simply served as a repository for results and then sent them downstream. “We’re not just implementing SCC’s product, but also extending it significantly, to have advanced real-time tracking in molecular labs built into the LIS, similar to what is traditionally seen in the LIMS [laboratory information management systems] that are used in research.”

Because the system will track the provenance and the quality of data generated in intermediate testing phases, recovery from quality-related problems will be more efficient. “Should we discover that there was a batch failure of master-mixed reagents or of one of our probe sets, after the fact or at any intermediate testing phases, essentially we can computationally excise all the questionable data, and generate an automated worklist for only that data that needs to be regenerated in support of the final result.”

Putting decision support at every workflow node within anatomic pathology is in the plan, Dr. Balis says. “We’ll generate bar codes at grossing, and throughout tissue processing we’ll have real-time knowledge of slides, blocks, cassettes, the tissue specimens themselves, and requisitions, and we’ll know their physical locations. So if a cassette goes missing, proactive alarms are triggered and that greatly constrains the area of search that will be needed to find it.”

What happens now is that a day or two goes by before the pathologist notices a slide is missing. “Then there’s an effort to search every possible location when maybe it’s not even known whether the block existed initially. A lot of the unknowns go away with real-time tracking,” he says.

Dr. Balis envisions a continued merging of AP and CP. “I think there will be specialized features of each particular workflow. But if you look at the roadmap that Soft Computer put in place, they don’t describe molecular or AP ‘modules’; they describe workflow engines, a very sophisticated repertoire of modules and steps that can be concatenated to build an overall, unified workflow that includes quality assurance, text manipulation, data storage, data transformation, data exchange—all the major elements of a relational database in an encapsulated data model.”

Not being information scientists for the most part, pathologists may find some of the concepts alien, Dr. Balis says. “At first blush, the challenge facing us is having a critical mass of laboratorian-informaticists who could participate in building the workflow along with the vendor. The tools are so sophisticated now that they can exceed users’ ability to maximally leverage them. And the vendors are in a race right now to add more functionality and features, so the challenge for the specialty will be keeping up with that.”

How to properly display AP reports in electronic databases is a chronic issue. “I have my own pejorative term for that process: the HL7 shredder,” Dr. Balis says. “With HL7-based databases, we lose control of the formatting, so the columns and page breaks are different, and one of the most egregious examples of unanticipated consequences is serendipitous breaks in the text where ‘no evidence of malignancy’ could be split so you would just have ‘malignancy’ at the top of the second page.” Such glitches underscore the need to “close the loop,” he adds. “It’s not enough to send reports to a downstream system and wipe your hands of it and say ‘we’re done.’ You need to communicate with clinicians, and make sure they received results and fully understand the significance of the results.”

The current generation of AP information systems is no longer just producing one report, in his view. “We’re producing two reports: a human-readable layer and then the machine-readable metadata layer, typically XML. So it’s getting more and more challenging to find ways to transparently generate that XML layer without bogging down the pathologist’s workflow.”

An IT issue that needs more attention in pathology is the proper management of consent forms, Dr. Balis says. Various universities have an aggressive opt-out consent form that assumes consent unless a patient opts out. “But in discussions with UM bioethicists, I’ve been assured that that model is not generalizable and is probably wrong. So I think in the long term, for any procedure we want to use, we should proactively seek permission for it.” From an IT standpoint, that means another metadata layer. “Our division is putting together multi-planar reports of narrative prose and then XML, and one can envision yet another metadata layer, which would be the XML that supports consent on a per-item or per-specimen basis.”

Challenges also remain for the handling of proteomics data, he says. “That falls under the general rubric of a high-throughput section within the clinical lab. You have massively parallel primary data, some need for meta-analysis, and then the ability to send out a multi-media final report that has graphics, images, text, and sophisticated formatting. And this is where the HL7 interface becomes a real problem. It’s essentially impossible to send one of those types of reports across intact without losing meaning.”

Collaboration between universities and vendors, such as that between UM and Soft Computer, used to be more common 20 or 25 years ago when the first generation of large automated information systems were put together, Dr. Balis says, noting that before he arrived at the university it was the alpha site for creation of Cerner PathNet, and other hospital systems such as Henry Ford and Massachusetts General have been working successfully with Sunquest. “But by and large, LISs have become something of a commodity purchase, and in that happening there’s been reduced opportunities for co-development.”

The vendors are not always on target in their innovations, in his view. “We’ve seen applications presented to us at USCAP by vendors that are way off the mark in terms of what’s actually needed at the bench.” Partnerships between “live” environments and the vendor development teams, for that reason, can be useful reality checks. A few other partnerships that he is aware of have been problematic even to the point of contracts being terminated, however. “So the results have ranged across the board from success to complete failure.”

A perennial topic at the pathology informatics meetings each year is open-source solutions in AP, which Dr. Balis says are achievable. But support over the long term is the main obstacle. “The problem with an open-source model is that even if it were perfect when designed, it would very quickly become obsolete. Implementation isn’t the challenge, but longevity is, because it requires a constant evolution, correction of errata, bugs, modification of workflow, and so on. It could work, however, if the informaticists and vendors collaborated on it. Then different vendors, as in the Linux model, would differentiate themselves by customization and the level of service they provide on top of the core platform.”

The universal standardized electronic medical record, on the other hand, will happen, Dr. Balis is certain, incrementally but inevitably. Like radiology with its DICOM standard, pathology will soon have interoperable lab results through an initiative termed LIDDEx (Laboratory Information Digital Data Exchange). At the University of Michigan, it’s just emerging from the demonstration stage. “When you have a digital interchange model based on cloud computing, it becomes easy to receive constitutive inbound variables, which are lab values, and send out calculated results, in this case for thiopurine sensitivity.” That project is about to go live, he says. “The architecture is working; it’s a matter now of just extending it to the outside world and finding clients.”

Traditional anatomic pathology systems had their own database with patient demographics and visit information exclusive to AP, says Gilbert Hakim, founder and CEO, SCC Soft Computer. But the need for integration “led us to create a single patient database going across clinical, AP, and genetics information systems, so the patient information and the status of the specimen are shared across all these modules.”

Three important changes have taken place in the past five years in AP, Hakim notes: increasing use of synoptic reporting, voice recognition and voice navigation, and digital pathology. “These three items are getting a major foothold in new pathology systems.”

Integration between AP systems like Soft’s and electronic health records can hit rough pavement at times, Hakim says. “AP systems may include graphs, charts, and clinical pathology results. For example, in a case of prostate cancer, you have drawings of the locations where the cancer may have appeared, you have a graph of the PSA that shows results over, say, four or five years, and you maybe have a picture of a slide. So the outcome of the report is a very rich-text PDF that the majority of EMRs cannot accept.”

Soft gets around this through a hyperlink. “We have an internal imaging system where we store the copy of this PDF and we send a hyperlink in HL7 to the major vendor EMRs. When the doctor clicks on that hyperlink in the EMR, the system automatically taps into our imaging system and displays the PDF in its original form.”

Discrete results are sent as well, in case clinicians wish to do their own trending. “But in general,” Hakim says, “the EMR systems are not advanced enough to accept the discrete data and reports of the colorized images through an XML form. There aren’t too many EMRs that can take advantage of this, so PDFs are an intermediate step.” He expects slow progress in EMRs’ ability to accept XML-based results. Much proprietary technology has evolved, keeping vendors from communicating with each other to address the problem. But in its partnership with the University of Michigan, Soft has already successfully tested the university’s use of cloud-based computing to have interchangeable documents as well as information traded across institutions with disparate information systems.

How likely is it that pathologists in five years will no longer have to navigate around those stacks of paper surrounding their desks? Hakim points to the advent of two-dimensional bar codes for all slides and cassettes, as well as increasingly error-free voice recognition that eliminates transcription, as tools that, along with all the other computing advances, will over time and ultimately make systems paperless. “All the technologies are there, and as the new generations of pathologists are coming in to use these new applications and the older generation retires, you’ll see this transition occurring.”

But such a trend will in no way phase out histology, Dr. Balis says. “There are some who are saying histology will become irrelevant because we’ll just put things in a test tube, grind them up, and with molecular next-generation sequencing and deep proteomics, you’ll have the whole story. But that doesn’t get at the heart of what’s going on at the structural level. Whether we use whole-slide scanners or microscopes, the power of microscopic interpretation shouldn’t be discounted. As a tool for rendering diagnostic information, I think the slide will continue to be incredibly powerful—whether we’re doing it with a microscope or a computer screen.”