On Point
Medical Mistakes: Human Error or System Failure

by Bill Bornstein

The now famous report on medical errors from the Institute of Medicine (IOM) - "To Err is Human: Building a Safer Health System" - included dramatic and now often-quoted statistics. At least 44,000 to 98,000 deaths may occur annually as a result of medical errors in US hospitals. These numbers, if accurate, would make hospitals the eighth leading cause of death in America and do not even include medical errors in the outpatient setting. This would rank the lethality of US hospitals ahead of motor vehicle accidents, breast cancer, and AIDS.

Very little controversy greeted the IOM report when it was released last November. To a large extent the initial reaction was, "yes, we can and should strive to reduce errors." The accuracy of the numbers was viewed as perhaps not so important. The areas of greatest controversy continue to center on error reporting, especially liability protection and whether error reporting should be mandatory or voluntary.

In the months since the release of the IOM report, many questions about the data have arisen. How accurate are the data? What exactly constitutes a medical "error?" Is the death rate due to medical errors really analogous to a large jetliner falling out of the sky every day or two?

Understanding the data allows us to put both the report and its criticisms in perspective.

Two studies form the nucleus of the IOM report. Interestingly, neither study is new. The 1991 Harvard Medical Practice Study reviewed hospitalizations in 1984 and identified 98,000 deaths related to errors. A 1992 study in Colorado and Utah extrapolated 44,000 annual deaths. In these two studies, respectively, 3.7% and 2.9% of hospital admissions incurred an "adverse event," or, in other words, an event resulting from medical care rather than from the underlying disease. The Harvard study identified 19% of the adverse events as drug related, while 14% were wound infections and 13%, technical complications. Overall, 58% of the adverse events identified in this study were labeled "preventable."

Both core studies used retrospective chart review to identify errors. It is clear that errors are underreported and that such review is necessary to ascertain the true rate of "preventable adverse events." This latter term is used synonymously with medical error in the IOM report.

Internist Troyen Brennan of the Brigham and Women's Hospital in Boston, one of the authors of the core studies that form the basis of the IOM report, thinks that "preventable adverse event" is a more accurate term than error to represent what those two studies explored. He points out in a recent article in the New England Journal of Medicine that preventability is often a subjective concept and that, in most cases, direct evidence of actual errors was not seen in the chart. Rather, preventability was inferred from the nature of the event. There is also an important link to cost. For example, are adverse events that are preventable at enormous cost truly preventable?

The distinction between error and preventable adverse event may seem insignificant, but it is not. For example, investigators in these studies assumed that many post-operative hemorrhages and wound infections could have been avoided with more scrupulous care. They classified such events as preventable without actually observing an error per se. This is probably very different from what the public construes to be an error.

There are also questions about the accuracy of the numbers. In July, the Journal of the American Medical Association ran point-counterpoint articles on the IOM report. Clement McDonald and colleagues of the Regenstrief Institute of the Indiana University Center for Aging Research argue that the number of deaths attributed by the IOM report to "preventable adverse events" is exaggerated. In his rebuttal, however, Lucian Leape of the Harvard School of Public Health contends that the figures probably underestimate the problem.

Despite these questions, the IOM report itself is largely constructive rather than critical. The report emphasizes that the cause of most of these events is neither negligence nor carelessness but, rather, the result of the inevitability of human errors. The theme is not that we must "do better" as individuals but rather that we must acknowledge our individual fallibility and implement systemic approaches to reducing and intercepting errors. As Leape puts it, "errors result from faulty systems not from faulty people…."

I think that most of us in the health care provider community recognize in this the clear ring of truth. We may not have been quite ready for this collective epiphany 10 or even five years ago. This fortuitous timing, in addition to the prestige of the IOM, may explain why the report is having such profound impact.

Error by any other name?

An important goal is to accelerate
development of a culture in which
anyone can bring up questions of
safety without fear of criticism, says
Emory Hospitals' associate administrator
and chief quality officer Bill Bornstein.

Historically, the operative model in the US health care system has been one of the provider as infallible superhuman. Unusual hours of continuous work during and after training have been part of the price of admission and a badge of honor. Health care providers have generally ignored human performance studies, viewing them as being applicable to more mortal populations.

Meanwhile, medicine has become exponentially more complex, in both its scientific underpinnings and escalating administrative burdens superimposed by a nation struggling with how to finance its health care. Until recently, American society has eagerly accepted our model of infallibility almost to the point of questioning natural mortality itself. Huge liability has been assigned to what, in other settings, might be viewed as simple, inevitable errors.

Today, health care organizations are increasingly adopting systematic approaches to quality improvement. Though coming to health care later than to many other industries, the methods of systems gurus such as Edwards Deming and Joseph Juran are being widely embraced. The crux of the message is that better processes, not greater individual efforts, produce the greatest enhancements of quality and productivity.

Acceptance of this concept represents a sea change in attitudes within health care organizations and has created an atmosphere fortuitously receptive to the IOM report.

The issue of reporting and liability is an important one. All agree that medical errors are vastly underreported, primarily because of liability concerns. Improved reporting of errors is not likely to occur without liability reform or at least protection. Furthermore, given the lack of excellent baseline data, improved reporting of errors will initially result in an apparent increase in errors.

Nevertheless, as Brennan points out in his New England Journal article, medical errors have been declining in recent years, especially in areas such as anesthesia. For some time now, anesthesiologists have focused on simple but important measures to reduce errors -- gas fittings that prevent the connection of wrong gases, consistency among anesthesia machines as to the direction in which controls operate, and work hours based on human performance data, to name a few. The death rate from anesthesia has fallen progressively from one in 3,000 to 4,000 in the 1950s to an estimated current rate of one in 200,000 to 300,000. To be sure, much of this progress is due to improved technology. However, experts in the field agree that simple system-redesign measures such as those just described may deserve most of the credit.

Finding a better way

Reprinted with permission of
Parade magazine.

What are the implications of all this for Emory Healthcare?

First, Emory Healthcare leadership is deeply committed to patient safety and minimizing medical errors. The components of Emory Healthcare were early adopters of system redesign for process improvement and have been using these techniques for the past decade or so. Second, although we have focused on improving patient safety and reducing preventable adverse events for some time, the IOM report is helping us increase awareness of these efforts and accelerate their pace.

Full and accurate reporting of preventable adverse events is an important element of error-reduction programs. Only by knowing the rate of such events can we measure the effectiveness of our error-reduction efforts. Furthermore, analysis of these events gives us the opportunity to introduce system changes to prevent their recurrence. As one authority in this field has elegantly put it, each report of a preventable adverse event is a potential treasure.

Emory Hospitals use a number of strategies to both prevent and quantify errors. We strive to create a nonpunitive culture in which reports of such events are truly valued. Discharge diagnoses are scanned for clues as to why preventable adverse events occurred. The pharmacy follows medications that are sometimes used to treat adverse events. Such medications include those prescribed to treat allergic drug reactions and those that reverse the effects of narcotic pain relievers and certain sedatives. Readmission diagnoses are scanned for evidence of hospital-acquired infections or other possible adverse events.

Despite these efforts, no health care organization knows the precise rates of preventable adverse events. Painstaking chart review, like that done in the Harvard Medical Practice Study and the Colorado and Utah hospital study, is likely to remain the gold standard for the foreseeable future, and, as already pointed out, even this approach has its limitations. Moreover, this type of chart review is highly resource consumptive and impractical on a wide scale. For the present, we must focus on putting in place systems of known benefit rather than waiting to be guided by precise metrics of error rates.

For almost 10 years, Emory Hospitals have used clinical pathways to help reduce preventable adverse events, such as pressure sores and postoperative blood clots. Protocols can help prevent these events.

Medication errors are one of the most common types of preventable adverse events. Emory Hospitals already have in place many of the medication safety measures recommended by the IOM. Among these measures are the use of unit dosing, central pharmacy supply of high-risk medications, and pharmaceutical decision support. We even use a robot (affectionately called Hal) in the pharmacy to dispense bar-coded medications into carts before they are taken to the floor.

Nevertheless, because of the inherent high risks of the medication process, we are further increasing our emphasis on medication safety. This year, Emory Hospitals have designated medication safety as one of their prime performance improvement focus areas. Each department is expected to develop a performance-improvement project in this area. In addition, the pharmacy and therapeutics (P&T) committees of Crawford Long Hospital and Emory University Hospital (which also serves as the Emory Clinic P&T committee) meet jointly on a regular basis to focus on medication safety. Wesley Woods has recently been invited to join this effort.

The joint P&T group is dissecting the medication process from physician's order to administration of the medication by the nurse. We know from the literature that the majority of errors in this process occur at the physician-ordering step. Fortunately, there are multiple downstream opportunities for these errors to be intercepted by the pharmacist, pharmacy information system, or nurse, for example. Of course, new errors may also arise during these intermediate steps.

We are addressing the physician-ordering step first and have agreed on a small number of basic standards for clear and acceptable orders. These include a clear and legible signature and contact number to facilitate contacting the physician for clarification when a potential error is intercepted by the downstream safety net. This is one of those trivial sounding measures that can have dramatic impact and extend beyond medication orders to all orders. Other elements of a minimally acceptable order include avoiding all but a few standard abbreviations and special rules for using zeros and decimal points.

We implemented these ordering standards in July to coincide with the arrival and orientation of new residents. In parallel, a series of educational initiatives are targeted at existing residents and attending physicians.

We also are increasing use of well-designed, dedicated medication order forms. Emory Hospitals are rolling out a new antibiotic order form in parallel with the new ordering standards.

Closer to home

What is a medical 'error?' How accurate are the data?

Better processes, not greater individual efforts, produce the greatest enhancements of quality and productivity.

One of our most important goals is to accelerate development of a culture of safety. Because errors in the ordering process are inevitable, questioning orders needs to be encouraged and gratefully received. In the Disney culture, they call it "stopping the ride." When a question of safety arises, any employee can stop a ride without fear of criticism. As the IOM points out, it is essential that we "make patient safety everyone's responsibility."

In the long run, information technology will go a long way to help reduce medical errors and increase patient safety. For example, optimally designed and implemented interactive direct physician order-entry systems with real-time decision support will have enormous beneficial effects on quality, safety, efficiency, and costs of care. Such systems are presently extremely expensive, challenging to implement, and not really available "off the shelf." Poorly designed and implemented systems can introduce new sets of risks and problems.

Stopping the ride

This is a rapidly evolving field, and we anticipate declining costs and more ease of implementation. Optimally timing our adoption of such systems within Emory Healthcare will be crucial. In the meantime, there are plenty of opportunities to optimize manual processes to enhance safety.

Despite the major challenges currently confronting the American health care system, I believe that we are on the verge of a "golden age" of health care. The increase in diagnostic and therapeutic tools over the past two to three decades is staggering. However, the enormous knowledge base now underpinning the practice of medicine has contributed to present discomforts. Only by aligning our understanding of human performance, process optimization, and information technology, can we truly deliver safe medical care and fulfill the promise of this unprecedented expansion of medical knowledge and tools.

Endocrinologist Bill Bornstein is associate administrator and chief quality officer for Emory Hospitals.

For the full text of the articles mentioned in this "On Point," see the NEJM Brennan article, and the JAMA McDonald et al. and Leape articles (AMA membership or JAMA subscription required).

In this Issue

From the Director  /  Letters

Imitation of Nature

A Cut, a Shave, and a
Blood Pressure Check

Medical Mistakes:
Human Error or System Failure?

Moving Forward  /  Noteworthy

Putting on the Ritz, Part Two



Copyright © Emory University, 2000. All Rights Reserved.
Send comments to the Editors.
Web version by Jaime Henriquez.