College of American Pathologists
Printable Version

  Informatics to the rescue in fixing ‘blunt end’
  AP errors


CAP Today




July 2008
Feature Story

In the wrong hands, “innovation” and “collaboration” can become buzzwords of the kind that put an audience to sleep. But when Jeffrey L. Myers, MD, describes partnering with informatics professionals for patient safety in anatomic pathology, people sit up and take notice. At Lab InfoTech Summit 2008 this past spring, Dr. Myers described the team approach that has taken hold at the University of Michigan Medical School, Ann Arbor, where he is A. James French professor of diagnostic pathology and director, Division of Anatomic Pathology. He explains how surgical pathologists and informatics professionals can collaborate to promote quality, safety, and efficiency. An edited transcript follows.

Let’s start with safety in surgical pathology. The medical community became engaged in a big way in the conversation about patient safety in November 1999, when the Committee on Quality of Health Care in America published its first report, “To Err Is Human: Building a Safer Health System.” The authors of this report said their goal was to break the cycle of your and my inaction. They pointed out that there would be no excuses despite the cost pressures, liability constraints, resistance to change, and other seemingly insurmountable barriers, technical or otherwise. It is simply not acceptable, they said, for patients to be harmed by the very system that is supposed to offer healing and comfort.

That was followed in March 2001 by the committee’s second report, which served notice that safety would not come at the cost of the other attributes of quality in health care. And they said care should be not only safe but also effective (defined as avoiding underuse and overuse), patient centered (something that’s rather foreign in the world of surgical pathology), timely (also foreign in some practices), efficient, and equitable.

But our focus here is what it means to be safe in health care, and safety in health care means being error-free. There are a lot of ways to think about errors in medicine. Lucian Leape, MD [Qual Rev Bull. 1993;19:144–149], categorizes them in four buckets: errors of diagnosis, errors of treatment, errors of prevention, and miscellaneous categories of error. Pathology could have connections to all four, but the two in which we chiefly reside are errors of diagnosis and miscellaneous categories of error. Leape says errors in diagnosis have multiple facets. There are errors in diagnosis (you got it wrong), but he also highlights—and he’s not speaking about pathology but I think we are the intended recipients—that delays in diagnosis are an equally egregious category of diagnostic error. It is not safe to make the right diagnosis too late to matter, and it’s not safe to fail to employ indicated tests (‘if we had only done the right thing we would have known the answer’), to use tests that are no longer the standard, and to fail to act on the results of our testing.

There’s been an interesting e-mail dialogue between academic leaders of anatomic and surgical pathology about the issue of critical values and reporting them in surgical pathology, and the tenor of the correspondence was, ‘This is not our problem.’ Can you imagine? Alerting physicians to critical values that will affect the lives of their patients is somebody else’s problem. We have a lot of work to do.

It’s easy to shrink away from some of the issues of error and say, ‘but we’re only human.’ And indeed, studies in multiple industries have concluded that when you look at accidents, 60 percent to 80 percent link to our human condition. Informatics can be key to overcoming that reality.

Errors have been personal to me for a while; they were personalized further in April 2007 when our laboratory made the front page of the Ann Arbor News for a mixup that led to an incorrect diagnosis of breast cancer. There we were, sharing space with the Tigers and the Pistons and drive-by monitors for car exhaust.

Our laboratory’s failure to accurately identify a specimen was no longer a secret. And it was said to be the fault of a laboratory technician, a human error that cost the university $3 million and a lot of public relations. The reporter’s implication was that if we had had a better technician, this might not have happened, but I have to tell you, it was our best technician. It was a talented and bright technician for whom this was so traumatic that she no longer works in health care.

Humans will fail. We have to figure out ways to avoid that for them and not leave them holding the bag. And we must take no comfort when they rather than us are named on the front page of the paper. The University of Michigan is not alone; I don’t know if you have seen your institution on the front page—lots of them have gained that notoriety. Last fall, newspapers and various Internet-based news media carried the story of Darrie Easen, who discovered only after having her double mastectomy that she never had cancer in the first place—another specimen mixup, this one at a lab in New York. A similar story originated in Minnesota earlier in the 2000s.

We’ve made a lot of popular press. The August 2006 issue of Readers Digest said in large font that pathologists are gambling with your life. And the article in that issue told the readers about floaters. Can you imagine? We hoped floaters would remain an in-house secret. Now patients know the many ways in which we can get it wrong. How often does it happen? Hard to measure; it depends again on how you define error in surgical pathology. If you define error as diagnostic discordance, which is a very insufficient surrogate for error, Andrew Renshaw, MD, and Edwin Gould, MD [Am J Clin Pathol. 2005; 124: 878– 882], and many others have shown that it can be as high as 11 percent and, interestingly, it varies depending on the driver for re-review of the case.

But Renshaw and Gould were different from other authors in that they linked that diagnostic discordance to the patient experience and pointed out that diagnostic discordance does not necessarily identify an error that is meaningful from the patient experience. If you look at historical data defining error more meaningfully as getting it wrong in a way that matters, meaning the error changed the way the patient might have been managed or changed the prognosis, then the number is closer to one percent or less. And it’s really these historical numbers in which we found comfort because historically that’s how we thought about quality in surgical pathology. We looked at these numbers and said, ‘Well, not so bad.’ At the University of Michigan, that error rate means 700 sentinel events a year. That will not play well on the front page of the Ann Arbor News. We’ve got a problem.

In quality assurance programs in surgical pathology and medicine in general, our thinking about errors has been focused on active errors—those that occur at the level of the front-line operator—because they’re easy to identify. Woods and associates [Behind Human Error: Cognitive Systems, Computers and Hindsight] described that as the sharp end of our complex health care system. It’s truly the interface between the practitioner and our monitoring processes, and therefore we focus by design on the active errors occurring at the hands of the practitioner who is left holding the bag. What we’ve learned since 1999 is that it’s an inappropriate way to look at error in surgical pathology or anything else. Our focus should be on latent errors, those that tend to be removed from the operator’s direct control. They reflect poor design and poorly structured organizations, and they’re much tougher to measure and resolve than are those front-line mistakes. And that’s why we’ve avoided them. It is, as Woods and associates say, the blunt end of the complex health care system, and in many respects that’s our topic here. How do you organize and develop policies and procedures at the blunt end of health care delivery that affect safety?

Don Berwick, MD, of the Institute for Healthcare Improve­ment, says it this way: Quality is an attribute of systems, not people, in medicine, and every system is perfectly designed to achieve exactly the results it gets. If your error rate today is .5 percent, that’s likely what it will be next year and the year after that unless you change something, because you’re designed to deliver at that rate.

That’s where we were when we were talking about safety a year ago at this conference. We concluded that significant errors occur in anatomic pathology practice at a low but predictable frequency. It’s a consequence of our complex processes that set us up to fail. And we left agreeing that we could do better, but we left this question on the table: How?

Which brings us to this question: How can informatics play a role in making surgical pathology safer?

In its 1999 report, the Institute of Medicine did offer principles that can guide solutions and countermeasures for addressing safety in medicine, and they apply to pathology in spades. First is providing leadership. That we’re having this conversation at a national meeting says that leadership in the informatics community is ready to play its part. Traditionally, when we have talked about informatics and safety we’ve focused largely on the Institute of Medicine’s second principle: respecting human limits in process design. If you’ve got your best technologist mixing up specimens, it means there’s a problem with the process, not the technologist.

What are the sorts of things that the Institute of Medicine was talking about? Avoiding reliance on memory and vigilance. Using constraints and forcing functions. Key areas in which informatics solutions have focused. They’ve also been the key tools historically in our quality assurance programs. ‘If you could just remember to get it right next time.’ I’ll try. ‘If you could just be more vigilant next time and not screw up the specimens.’ I’ll try.

What informatics solutions are you familiar with? Provider order entry. Patient specimen identification—solutions that are much more robust than what we’ve had in the past. Things like standardized workflow driven by computerized workflow models and standardized results reporting. Templates that are now common in many practices. And case-appropriate content delivered to the practitioner at the time of signout that might inform the final product before it’s verified and distributed. The Institute of Medicine also advised simplifying and standardizing your processes, areas in which informatics can play a role.

The Institute of Medicine talked about other areas in which it’s harder to see the link to informatics, but they have everything to do with what you are about: promoting effective team functioning, anticipating the unexpected, and creating a learning environment around safety and quality. And safety and quality are related no matter how you define quality.

Data collected and published recently showed that one of the distinguishing features of the organizations with the best safety records in health care was that their leaders were focused on service. They are not unrelated. And it’s about creating a learning environment where you celebrate errors as opportunities to be better. It’s about encouraging the reporting of errors and doing that in an environment in which there are no reprisals. Again, this is an area in which informatics has traditionally had a presence. Thomas and Petersen [J Gen Intern Med. 2003;18:67–67] suggested that if you look at the various ways to measure error in medicine, you will see that some get you closer than others to the blunt end of our complex health care system, and IT, in their view, is one of those ways. The role of informatics historically has been to do things like identify events, generate reports, and perhaps provide data analysis.

Some of the other conversation around creating a learning environment doesn’t look like it’s linked to informatics, but I’m going to show you how I think it is. They said you have to develop a working culture in which communication flows freely regardless of authority gradient, something that medicine learned from the airlines. What they discovered that was fundamental to improving the safety of flight in the ’70s was that it was unsafe to fly in a culture where your copilot was afraid to tell the pilot that they were going down. Where the copilot was afraid to tell the pilot, ‘Did you notice the large 747 at the end of the runway?’ Conversation never occurred and it was a disaster. And medicine’s no different.

Creating a learning environment is about implementing mechanisms of feedback and learning from our errors. The perception is that we’re not so good—at least from those outside of health care. In an interesting survey published in 2002 [N Engl J Med. 347:1933–1940], Blendon and associates surveyed over 2,000 respondents, some of whom were physicians and some of whom were nonphysician members of the public at large. They asked about how often they experienced error; it was a high number, 40 percent. They asked how often it was serious; it was a high number, 10 percent. They asked how serious; the answer was ‘pretty serious’—we’re talking about death here. And when they said, ‘What do you think the very important causes of that experience are?’ the public had as No. 3 on its list, ‘You guys are lousy at teamwork.’ Failure to work together or communicate as a team. Interestingly, the physician respondents did not see it that way; it was much lower on their list.

So the challenge we’re talking about is, How do you create meaningful organizational links, to overcome that weakness, between surgical pathology and informatics, and how are we going to partner to facilitate innovative solutions around safety?

That takes us to the question, How do you align surgical pathology with informatics resources?

First, the principle. In my view, there are three imperatives around this alignment. The first is alignment at organizational and administrative levels. Where are you on your organizational chart, if you will? That will tell me how you work together. Why should that matter? It matters because this can be successful only if informatics and the operators in surgical pathology have a common vision and mission rather than visions and missions that overlap, or worse, are diametrically opposed. They have to have common strategic goals, and they have to have a common understanding of operational imperatives. Why? Because that’s where you decide to spend your money.

How does organizational and administrative alignment happen at the University of Michigan? I’m sure there are a lot of ways to do this successfully. At the University of Michigan (where it happens in a manner that’s successful) informatics is one of seven coequal divisions in the pathology department. It has both clinical and research arms; Ulysses J. Balis, MD, is the current director of clinical informatics. The closest neighbor to informatics—at least in terms of the clinical enterprise—has always been clinical pathology. And I would imagine that’s because with the rate at which CP created demand with new testing modalities of the last few decades, it was the primary consumer of informatics resources. When you talked about meaningful links at a clinical level between informatics and practitioners of pathology, this was the link that mattered. Which results in some odd conversations that irritate someone who does what I do for a living.

In fact, I’ll share with you the No. 1 on my top 10 list of things not to say to people like me when it comes to informatics. It’s something I commonly hear on site visits when we’re trying to imagine what the future of informatics will look like in surgical pathology: ‘Our LIS was chosen by CP. It actually doesn’t work well for us.’

When I hear that I have a lot of questions about our partnership and what it’s been historically. I would argue that anatomic pathology is creating demand for informatics services that has either overtaken or will soon overtake CP. Why is that? It’s things like image digitization and its role in future practices. It’s about new testing and research technologies that blur the division between surgical pathology and clinical pathology. And one of the key drivers is recognizing how bad we are in delivering safe, high-quality care. That’s what’s generating demand for informatics solutions, and that’s why we’re talking about anatomic pathology at Lab InfoTech 2008. So the relationship that makes sense to me is one in which AP, CP, and informatics are aligned, as they are at the University of Michigan. Organizational alignment is key to getting it right. If you live too far apart, if you don’t share a common vision and mission, if you don’t have a common purpose that drives resource allocation, then it’s hard to get it right.

The second imperative is alignment of tasks. By that I mean a deep common understanding of practice needs and the resolve to respond with innovative and timely solutions. And that’s easily said and very difficult to achieve.

The way we’ve taken a run at that at the University of Michigan has to do with the way we’re organized around our laboratory operations. The supervisors and lead technologists in the various AP laboratories have a direct reporting relationship to chief technologists, who report to our dedicated operations administrator, who works closely with me. Our chief technologists have a dotted-line reporting relationship to our medical directors, but everyone is engaged in the central AP Laboratory Operations Group. This is not because I’m a fan of bureaucracy; I’m not. It has to do with a Lean view of how we work. The charge for this group is to provide administrative oversight of laboratory operations, which for us means specimen receiving and accessioning, which also captures our intraoperative frozen section practice. It has to do with cytopathology, histology, immunohistochemistry, electron microscopy, autopsy, and, importantly, AP-related laboratory information systems. The reason you need a group that provides oversight across that spectrum is because our value stream plows through all of those silos. We have to make sure someone has the big picture as we’re making resource decisions in individual laboratories. Our purpose is to identify and prioritize our opportunities for enhancing and improving our performance using Lean tools and principles. This is the group that’s driving the adoption of Lean thinking in AP at the university. It includes clinical informatics leaders and our laboratory managers, our laboratory directors, and physician leaders from not only anatomic pathology but also hematopathology, which organizationally resides elsewhere in our department. It includes members of our administration and some of our trainees.

The third imperative for aligning surgical pathology and informatics is alignment of strategies, by which I mean you have to be committed to a common strategic vision and have a shared definition of excellence. This means we’re both going to measure our success using the same metrics. There are places that not only don’t have a shared strategy but actually have competing strategies.

What’s our shared strategic vision at the University of Michigan? Our goal is to be top of mind when anyone anywhere imagines excellence in applying clinical informatics to solutions in anatomic pathology. If we can do that, both of us have won.

That’s the principle. What’s the experience? I’m going to share with you the experience in the last two-and-a-half years because that’s how long I’ve been at the University of Michigan.

We’ve developed new tools with a focus on safety. One is our critical values alerts management report, a tool that daily mines our database to discover cases that look like they might be critical values but haven’t been communicated in a way that is consistent with our policy. We use that for two things: to measure our own compliance with our policy and use those events as educational opportunities and to intervene real time in patient care. If a result hasn’t been communicated that should have been, I get on the phone. This is a report that I look at daily, or that Dan Visscher, MD, our director of surgical pathology, looks at daily, so when something has fallen through the cracks, we feed it back to the practitioner who failed to do the communication in the hope they’ll do it better next time. And then we say, ‘Do you want to call or should I, because somebody’s going to call.’

We’ve also, with Dr. Balis’ help, developed unique applications, and I’m going to talk about one for facilitating case handoff. When I arrived at the university, our practice was such that patients were not accessioned to people but to services, which was a problem. Why? Because if something falls through the cracks, you don’t know whom to call or whom to talk to. And when you go to the room where the service lives, mostly you’ll just get blank looks. So we developed a policy that created the expectation that a case can change hands only with faculty-to-faculty communication. One of the ways we can do this is by using a Web-based form that we call a Situation/ Background/ Assessment/ Recommendation/ Cali­brate form, or SBARC. It’s borrowed from the military, minus the ‘C’ for calibrate. This is a communication principle developed in the Navy where handoffs in terms of communication can be pretty important. It was adopted at the University of Michigan in nursing, which collaborated with the business school to understand how this might work, and they added the ‘C,’ and we use it now in pathology. We had to have a vehicle to facilitate the communication and it had to be easy. So what we have is a Web-based form in which you enter a little bit of text in each of the boxes. It doesn’t have to be much—a word or two. Pull-down menus are used to send it to the receiving faculty member, and both faculty members get a copy in their e-mail boxes.

Other tools we’ve developed, some of which many other labs have: bar-code–driven electronic order entry for special stains, bar-code–driven block labeling, and a unique stalled case management report that’s meant to facilitate flow in the sense that Lean defines flow. When cases sit still for too long, you’re at risk for a variety of reasons. So this is a report, generated daily, that identifies cases that have inexplicably stalled at various checkpoints along the process. It was grossed yesterday but nothing else has happened—that’s out of bounds, what’s happening? It’s been a powerful and useful tool to ensure flow in our practice.

I also want to tell you a story about the redesign of our surgical pathology rotations that was implemented in August 2007. This was meant to get at safer, higher-quality practice.

This began with a design phase in January 2007 and went live in August. The tools to facilitate this transition were built at the level of the AP Laboratory Operations Group, so informatics was at the table. We did a lot of things in anticipation of this dramatic change in our practice, but much of it had to do with creating tools that would support flow. In addition to the tools I already described, we built important Web-based calendars. Now who doesn’t have Web-based calendars? No big deal. But they’re more than just a place to go to get a name.

What I needed was a calendar that sent a message about expectations for team participation, which was a big cultural shift at the University of Michigan in August 2007. We now have team-­based calendars where you don’t learn the name of just the attending on a specific service. You also learn the names of the resident, fellow, and, very important for us, the PA. PAs were kind of faceless and nameless in our practice before, and we made the decision that they were going to be key participants in this service transition. I needed more than the names; I needed a way that made it easy to go from the names to the people. This tool allowed us to click on one of these names and immediately come to the window that makes it possible to do an alphanumeric page. It helps us communicate. So now nearly every day that I’m on a service I say to the resident, ‘Who’s the PA today?’ and if they can’t tell me, we have words.

This is a tool that speaks to creating a learning environment. It’s about developing a working culture in which communication flows freely regardless of authority gradient. Our colleagues in informatics were key to developing the tools, and they have been key participants in our Lean projects.

I will close by sharing some of the other unexpec­ted opportunities that emerge when surgical pathology and informatics are aligned. This opportunity surfaced in a phone call with a remarkable three-person practice in Venice, Fla.: Richard Whisnant, MD, Tammy Densmore, MD, and Bill Roth, MD. Richard trained in Pittsburgh, Tammy and Bill at Washington University in St. Louis. Their practice is called Suncoast Pathology, and they had built an out-of-the-box informatics solution to common problems in pathology practice.

They built elegant but simple solutions and came to us and said, ‘Can we do this in a way that functions at an enterprise level?’ We now have a relationship and are going forward with that. They developed things like bar-code–driven specimen identification and workflow monitoring at grossing, embedding, and cutting stations, and they can tell you in real time, just like a flight controller, the precise status of the operations in their labs at any given moment. They built gross descriptions that used template-driven pick-lists and free text, and they have a suite of unique tools to improve diagnostic performance while minimizing preventable workflow errors, things like bar-code–driven patient ID and workflow monitoring for the pathologist, not just the technologist.

They have interesting forced functions that, among other things, force the person signing out the case to acknowledge certain patient-specific historical data. They have a way of mining for data that might be relevant, they have tools that prevent accidental verification of incomplete cases, and they have tools that ensure compliance with QA principles to which they’ve all agreed but occasionally, like the rest of us, forget. They have template-driven reporting that’s customized by client. They have real-time tracking for diagnostic performance. They can tell, using rates of concordance as a measure, whether they’re getting better or worse or staying the same. And they created opportunities to pull relevant knowledge content linked to unique report elements at the time of report verification, such as: ‘Did you remember that this is how you grade endometrial cancer?’

Richard’s a techno-savvy guy, but his two colleagues would not describe themselves in that way. So when we went to Venice to hear about this, I thought, Well, I know what I’m going to hear. Richard is going to say it’s great and his two partners are going to say, in secret, quiet tones, ‘This really sucks. We’re slower now. We miss paper.’ It’s a paperless process. What they said was, ‘Oh my God, our life is so much better. We actually leave work on average two hours earlier having done a better job.’

My purpose was to talk about the problem of safety in surgical pathology, to talk at a high level about informatics’ role in seeking solutions to what is a big problem. And to leave you with what I hope is a positive message: that aligning surgical pathology with informatics resources is important and a fundamental principle of making health care safe in our environment. Our experience to date at the University of Michigan makes the point.