Creating a Safer Health Care System: Title and subTitle BreakFinding the Constraint

Primum non nocere. In Epidemics,¹ Hippocrates urged all physicians to provide safe and effective care. Of course, that is not always possible because adverse events or bad outcomes sometimes occur. But physicians are driven to minimize the likelihood of harm, especially adverse events that more careful reasoning, better practices, and better systems might prevent. The Institute of Medicine called attention to the ubiquitous nature of errors in medicine and issued the challenge to develop safer systems of care.^{2 - 4} Safety, the flip side of doing the right thing at the right time, means not doing the wrong thing at any time, especially if the wrong action occurs inadvertently—by mistake—because the clinician or the system did not prevent an error before its effect reached a patient and caused harm.

Now, 5 years after the Institute of Medicine report,² several authors have questioned the system's progress toward patient safety.^{5 - 8} In this issue of JAMA, Longo and colleagues⁹ provide data about the pace of implementing clinical safety systems. When a system had a relatively high rate of implementation initially, subsequent improvement was difficult¹⁰ ; when a safety system did not have high penetrance initially, more rapid change might be expected subsequently. There has been steady progress in patient safety, but it has been perhaps unexpectedly slow.

Why, then, are these important improvements taking so long? Is the problem inertia, inattention, a lack of resources, or conflicting demands on time or energy? Medicine is no longer an island; physicians need to learn from other domains of practice, other industries. Of course, no industry has a uniform record of success; they all must learn from each other.

Over the past 2 decades, a set of tools and concepts loosely called the theory of constraints^{11 - 15} evolved from an initial focus in manufacturing. The theory of constraints requires explicitly framing the desired goal, the necessary conditions for achieving that goal, and the measurements that will monitor progress toward that goal, and then establishing a clear plan for making change happen. For medicine, the goal is to improve health now and in the future. One of the necessary conditions for moving toward that goal is to make clinical care safer. Having identified an undesirable effect, ie, that errors occur at a substantial frequency, the key question is why does this undesirable effect persist? Why have safer systems of care not been implemented? From the simplest perspective, the theory of constraints asks 3 questions: (1) What should change? (2) To what should it change? and (3) How should change occur?

Most systems and most individuals resist change. Systems must have substantial inertia to make them stable, and medicine is no exception. In many ways, medicine is still a “cottage industry” of individuals (both clinical and administrative) who do things their own way, in their own silos. Only recently have physicians seriously considered systems of care and required students and trainees to learn about them. But just as Prochaska et al¹⁶ have described the stages of change in patients' behaviors, the theory of constraints teaches the layers of resistance to change in a system, and for each layer how to achieve “buy-in” so change can happen.

First, there must be agreement that there is a problem. Most clinicians likely agree that medical care is not as safe as it should be, although some might still disagree that errors occur frequently. Recognizing that clinical care can be unsafe shakes physicians' belief that they are doing all they can for patients and, certainly, that they are not harming their patients.

Second, there must be agreement on the direction of the solution, an area that still requires substantial work. As demonstrated by the extensive list of criteria described by Longo et al,⁹ physicians have not had sufficient focus. The likelihood of successfully impacting 91 measures of progress in patient safety is small. Even the mathematical construction of 7 latent variables, based on factor analysis, does not provide deep knowledge about why the medical system remains unsafe or identify the core problem that must be addressed. Longo et al⁹ seem to be trying to improve every step of the process of care, taking the philosophy of total quality management,^{17 - 19} which tries to reduce variation to a minimum everywhere. But in medicine, certain kinds of variation cannot be eliminated, such as biological variation and preference variation from patient to patient. In fact, the theory of constraints emphasizes that where variation cannot be eliminated, systems must be developed to buffer against it and manage it.

Third, there must be agreement that the proposed solution actually solves the problem. Here, considerable work remains to be done. Even for something as prominently emphasized (and as expensive) as computerized physician order entry, disagreement about its effects on safety and efficiency remain.²⁰

Fourth, there must be agreement that the solution does not introduce new adverse effects, so-called negative branches, or that any such branches can be successfully trimmed, shaping an acceptable solution. Ideally, such unintended consequences should be trimmed as the solution is designed, but the solution also can improve after a pilot. Those who omit this step often rediscover Murphy's law: “Nothing is ever so bad that it can't get worse.”¹³ Such negative branches occurred in the initial implementation of the Health Insurance Portability and Accountability Act²¹ and may well occur with widespread implementations of computerized physician order entry.²⁰ Careful planning and pilot implementations will be important. Here, too, there is work to do in patient safety.

Fifth, it is essential to identify the obstacles to implementation and agree on how they might be overcome. Again, clinicians have hardly begun to plan for change at this level of detail but, rather, struggle in a clinical world of limited resources, conflicting needs, and unfunded mandates. Absent major increases in funding, redirecting resources to improving patient safety likely means that other necessary capital and operational improvements will be delayed.

And, sixth, although change sometimes shakes core beliefs, physicians must agree to implement the change. Even though there is no national consensus about what to change, what to change to, and how to accomplish the change, the lack of focus and lack of planning contribute to the slow pace of change in patient safety.

Improving the safety of patient care must be a high priority for all clinicians and administrators; it must become a central part of their training. Making health care safer requires considering the forces, aside from inertia, that keep the pace of change slow. An aphorism ascribed to Deming, and restated by many medical authors, states, “Every system is perfectly designed to achieve the results it gets.” Physicians must have profound knowledge of the health care system to understand why it behaves as it does, and they must understand that “why” before reasonable change can occur. Deming argues that such knowledge must come, in part, from outside the system.¹⁸ In the terms of the theory of constraints, undesirable effects (in this case, adverse patient clinical events) persist because some underlying policies and measurements reinforce the behaviors that produce them. The theory of constraints teaches, “Tell me how you will measure me, and I will tell you how I will behave.” For example, on one hand, physicians seek to make care safer, while on the other hand, they also seek to make care more efficient and less costly. Because cost can be measured more precisely than quality or safety, cost often has a greater effect in shaping behavior than does quality or safety. To improve safety substantially, clinicians and managers must discover what policies and measurements are producing the behaviors that continue to make the system unsafe.

Do safety systems increase costs (at least in the short run) or require capital that is not available or is being deployed elsewhere? Will safety systems diminish throughput and take more of clinicians' most precious resource—time—and perhaps lead to more errors of omission? Have clinicians and administrators engaged all the stakeholders and gotten their buy-in? The complex processes of clinical care require communication, handoffs, and synchronization. The lack of synchronization of the current clinical workflow is a core problem; achieving such synchronization will require developing and using appropriate software.

Medical care is a chain of processes that together improve a patient's health. Each step can be associated with variation, failure, and even errors. Throughput is the rate at which a system moves toward its goal¹¹ or, in this case, the rate that it improves health. The theory of constraints teaches that every system has bottlenecks, steps where its throughput is most constrained. If a bottleneck is kept idle because of variation or a failure in the steps that lead to the bottleneck, the system cannot make up for such delays; throughput is lost forever. To manage variation properly and maintain throughput requires feeding the bottleneck though a buffer so that the chance of an idle bottleneck is minimized.¹¹ However, this key lesson from manufacturing does not often get translated to clinical systems. The goal of providing safe and effective patient care would be best served if clinicians and administrators understood the constraints to achieving that goal and buffered against variation upstream from those steps. In a world of limited resources and limited energy, the theory of constraints asks clinicians to focus and provides the tools to identify the core problem or core conflict, the point at which a change will have the greatest effect in moving toward the goal.

The theory of constraints provides 3 common measures of system performance: throughput, inventory, and operating expense.¹³ An effective system maximizes the former while minimizing the latter 2. But, not infrequently, actions that decrease inventory and operating expense also cause throughput to decline. For example, financially challenged health care systems sometimes cut expenses so severely that throughput (and perhaps length of stay) is adversely affected. Because inventory and operating expense have practical lower bounds and because throughput measures approach to the system's goal, throughput must be protected; it is the key measure, especially in not-for-profit organizations.¹³

To produce sustained change, it is essential to understand root causes of current problems, establish policies to induce and maintain change, create measurements at all levels that shape safer behaviors, and properly measure progress toward the goal of having a safer health care system. Longo et al⁹ provide data about the introduction of safety systems, but better measurement systems and better data are also needed about the incidence of adverse events.^{22 - 23}

Rewarding safety will surely help. Some clinicians might consider being paid to perform as being unprofessional, but few could object to creating a safer and higher-quality health care system. Rather than labeling such initiatives as pay-for-performance programs, it may be preferable to think of them as paying for quality and paying for safety. The time has come to take bold action and to embrace change, but first it is time to understand the constraints to accomplishing that change. As Deming¹⁸ said, “Change is not necessary; survival is not mandatory.”