Earlier papers in this series have explored the conceptual roots of modern approaches to improvement, which lie in systems theory [8]. A system is a collection of interdependent elements that interact to achieve a common purpose. Performance-measured in health care by such yardsticks as health outcomes, complication rates, waiting times, and patient satisfaction-is a system property inherent in the work of physicians. To improve systems, we must usually attend first and foremost to interactions. Among humans, "better interaction" is almost synonymous with "better cooperation." Those who aim to improve health care must be students and practitioners of cooperation, not because it is nice (although it often is nice) but rather because mutual aims require it.

Actually, cooperation is not always essential for improvement. The degree of interdependence in the system makes a difference. Take sports as an example. For a bowling team, achieving the highest score depends on every person's doing his or her best. Cooperation, except perhaps in exchanging tips, is not the primary issue in successful bowling. But a basketball team courts failure if every player's primary goal is simply to score the most possible personal points. Playmakers and rebounders make essential contributions to team success. More often than not, a balanced team that passes well will beat an equally talented team full of players who hog the ball.

Health care has its "bowling events," to be sure, especially when the lens zooms in on the individual encounter. Alone in the examination room with a grieving patient, the physician does his or her best for the patient primarily through individual skill and compassion. But zoom out again, and the basketball analogy seems stronger.

A single physician managing patients with chronic asthma may have difficulty developing systems that prevent exacerbations by reaching into patients' homes or training patients to monitor their own pulmonary function. But embed that physician in a group practice or a sound managed care organization, and responsive, willing, cooperative others-such as managers, nurses, pharmacists, and health educators-may be able to configure supports that help everyone, including the physician, the patients, their families-even payers-all at the same time.

Scientific Foundations for Successful Cooperation

A review of modern theories of the foundations, uses, and problems of cooperation is beyond the scope of this paper, but a few examples suggest the richness of the scientific terrain for potential students of the topic.

Ronald Coase, from the University of Chicago, won the Nobel Prize in Economics in 1991, largely for theoretical work on transaction costs closely related to normative theories of cooperation. In one of his most influential papers [9], Coase relates the story of a railroad with a right-of-way through a farmer's fields. As trains crossed the field, they spewed sparks, damaging crops close to the tracks. Coase asserted that the usual social approach to the problem, a "rights-based" search involving lawyers; lawsuits; litigation; and, probably, an enforced remedy under which the railroad might have to install equipment to suppress the sparks would not produce the solution with the lowest cost to society. Instead, if the railroad and the farmer could bargain easily, they would probably discover a far less costly solution; for example, the farmer might agree not to grow crops within reach of the sparks if the railroad would pay the farmer for profits thereby lost. Coase's vignette easily brings to mind any number of problems in health care (such as neurosurgeons and orthopedists arguing for operating room time) that often lead to "fairness" solutions (equal time) or resource investments (building more operating rooms) instead of solutions that are optimal for the system as a whole. We argue that this is evidence of a low level of cooperation between the specialties.

Coase asserts that leadership in organizations can provide a context for purpose-based conversation or bargaining instead of reliance on more costly, rights-based solutions. Reinertsen [10] describes the physician's role in providing this type of leadership and methods for executing those leadership responsibilities. However, it is difficult to identify someone who might provide that leadership. Hiatt [11] recognizes that some aspects of health care lack obvious leadership in fostering cooperation. In particular, he refers to three kinds of conflict over the use of resources for medical practices: conflict between the interest of the individual and the interest of society, conflict over whether a medical procedure provides value, and conflict between preventive and curative practices. Coase shows that lack of leadership can be expensive.

If this leadership is missing, is cooperation impossible? This question was addressed by Robert Axelrod, from the University of Michigan, who extensively explored the famous game theory problem, the Prisoner's Dilemma, in The Evolution of Cooperation [12]. Axelrod was primarily concerned with cooperation as a self-initiated strategy for an individual person working in a system rather than with the leadership-initiated cooperation explored by Coase. His work gives hope that the health care system can be positively affected by physicians skilled in cooperative behaviors.

In the Prisoner's Dilemma (Table 1), two players each face a matrix defining the consequences of their independent, blinded choices to "cooperate" or to "defect." If one player cooperates while the other defects, the defector wins. If both defect, both lose. If both cooperate, both win, but not as much as a lone defector does.

In the short run, defection is the only rational choice, assuming that each person seeks only to maximize his or her gains independently. To see why, assume that you are player 2 and work out the consequences of your own actions in a one-cycle Prisoner's Dilemma game. No matter what player 1 does, you are better off defecting. The game gets interesting and the ideal strategy gets less obvious when the game is continual or has an uncertain future. If player 2 is shrewd or powerful enough to maintain the result in the upper right-hand cell, only altruism would motivate player 2 to move to another cell. (Perhaps physicians and hospitals were in player 2's position in the past, when society was willing to pay double-digit annual percentage increases in health care costs.) However, continual refusal to cooperate usually leads to a system that equilibrates to the bottom right-hand cell, producing unsatisfactory results both for the two persons playing and for the system as a whole, like wasteful hospital construction in an over-bedded city.

Axelrod invited game theorists, economists, and mathematicians to write computer programs to play Prisoner's Dilemma, and when all entries played all other entries in a series of games, one strategy dominated. It did not win every match, but it came out ahead on the average across all matches. Axelrod dubbed the winning strategy tit-for-tat, and its algorithm consisted of only two rules: cooperate on the first move and, thereafter, do what your opponent did on the previous move.

In a similar vein, Fisher, Ury, and Patton [13], along with their colleagues in the Harvard Negotiation Project, have spent two decades developing and refining teachable approaches to what they term principled negotiation, strategies of conflict resolution. They aim to provide methods that can help parties invent new solutions and move beyond the static, constrained payoffs of the Prisoner's Dilemma. Their work emphasizes the importance of relationship in the quest for good solutions-best of all, "win-win" outcomes to disagreements initially thought to have only "win-lose" possibilities.

Applications of research findings on cooperation have resulted in improved safety, reliability, and effectiveness in many industries. Aviation safety experts, for example, have long known that communication failures account for more than three fourths of commercial airline accidents and that sociological barriers to communicating crucial information "against the authority gradient" (for example, from a navigator to a pilot) often lie at the root of these communication failures [14]. Building on this knowledge, the aviation industry has invested heavily in training crews (and especially their highest-status members, pilots) to ensure highly cooperative norms and behaviors regarding mutual assistance and exchange of information as mainstays of safety. This field is known as cockpit resource management [15]. Although extrapolation of technical methods from one industry to another requires caution, researchers in simulated operating rooms have found the same communication hazards found in aircraft cockpits. They suggest that cockpit resource management techniques may have as much promise in health care as in aviation [16].

Other writers and researchers have made equally important contributions to our understanding of the uses and methods of cooperation in health care and in other industries. We summarize these few here not to give them preeminence but rather to support our assertion that cooperation is a proper subject for study by physicians, a valuable topic for scientific investigation in health care research, and a proper arena for leadership actions toward improvement.

Methods for Fostering Cooperation

We suggest five scientifically grounded methods to foster cooperation: 1) develop a shared purpose; 2) create an open, safe environment; 3) include all those who share the common purpose and encourage diverse viewpoints; 4) learn how to negotiate agreement; and 5) insist on fairness and equity in applying rules.

We discuss these methods at the organizational level and at the level of the individual physician. The system-level application of the methods is described with the aid of one especially successful example of system-level cooperation in a care delivery site where interactions matter a great deal: the modern intensive care unit. At the individual level, we suggest actions that a physician can take to improve his or her cooperation skills even while he or she is working in an organization where cooperation is not the norm.

The Shock Trauma Respiratory Intensive Care Unit (STRICU) of the LDS Hospital in Salt Lake City, Utah, made a commitment in 1992 to transform practice and create an environment supportive of practice improvement. Physician staff in the 12-bed STRICU included four full-time critical care teaching attending physicians, four pulmonary-critical care specialists who consulted frequently, and more than 55 affiliated attending physicians who admitted patients under an open admissions policy. The open admissions policy made establishing a culture of cooperation particularly important. In addition, about 45 nurses, 8 respiratory therapists, 8 technicians, and 5 clerks worked in the unit. No single person was in a position to dictate policy or procedures.

In the first 4 years of the cooperative effort, significant gains were documented in the use of sedation and paralysis; in the control of serum glucose [17] and potassium levels and the frequency of brain edema; and in the use and costs of antibiotics [18], enteral feeding, laboratory tests, and pharmaceutical agents. During these 4 years, costs have decreased by as much as 30% in areas directly influenced by improvement teams, resulting in an overall 19% reduction in total hospital costs [19].

Develop a Shared Purpose

A system needs a purpose to help persons manage their interactions. In searching for the motive for cooperation, a stronger answer than "financial incentives" is "purpose." Systems, especially human systems, derive their identity from a shared, common purpose. Take away purpose, and the system dissolves; the interactions no longer have meaning, and one has only isolated elements. Cooperation, like a system itself, begins with a common purpose based on that which is perceived to be of greatest importance to the participants. The statement of purpose for a practice, a hospital, or an integrated delivery system should address patient outcomes, service, and total cost.

In the STRICU, the unit medical director and the nurse manager led an effort by front-line workers from all disciplines to develop a statement of purpose. This statement emphasized the importance of focusing on improving outcomes and working efficiently. It also provided guidance for interactions among clinicians and staff and became the centerpiece for a series of workshops designed to increase the abilities of all STRICU employees and thereby improve the performance of the unit and develop supportive working relationships.

An individual physician who is involved in a conflict or a situation requiring mutual decision making can foster cooperation by beginning the dialogue with a question (Table 2): What are we trying to accomplish? This question will provide an opportunity for understanding common aims and moving beyond initial positions.

Create an Open, Safe Environment

Cooperation can feel unsafe, especially at first. Physicians may create fear within groups and organizations. Nonphysician staff with good ideas or requests for change whisper those ideas and requests or withhold them entirely, fearing criticism or reprisal from the higher-status physicians. Similarly, physicians may underestimate the power they could have as role models if they were to engage in new forms of cooperative activity and be willing to give up something of their own to improve the effectiveness of the team.

In the STRICU, the medical director and the nurse manager took responsibility for building an environment conducive to cooperation. Initially, they modeled cooperation in their interactions while running the unit. Simultaneously, they sent signals to colleagues that direct, open communication among disciplines was desired. For example, when either of them was approached by a physician complaining about a nurse or a nurse complaining about a physician, they encouraged the person with the complaint to address the issue with the other person involved, and they taught them how to do so in productive, nonthreatening ways. The evolution of the cooperative environment was accelerated by multidisciplinary teams improving patient care (for example, by developing and using protocols). One of the first protocols, management of mechanical ventilation, was developed by a team of physicians. Later, nurses and respiratory therapists were included on the protocol team. Subsequently, all teams included representatives of multiple disciplines and, when appropriate, were led by someone other than a physician. For example, nurses led the teams that developed the sedation and potassium replacement protocols.

A physician can help others engage in cooperative behavior by being the first to model willingness to suggest and test a change (Table 2). For example, if a primary care physician thinks that changing his schedule for rounds could increase the chance that he will encounter a specialist and thereby smooth communication or improve the flow of patients in the hospital, the primary care physician should try the change and report back. If the change worries the physician, he should scale down the test but should not avoid action.

Include Those Who Share the Common Purpose and Encourage Diverse Viewpoints

Convening those working in various roles in the system establishes the opportunity for cooperation. Enough persons should be included to permit presentation and discussion of the aims, current performance, and boundaries of the relevant system. The physician's goal as leader of cooperation should be to seek, or to start to create, a shared sense of purpose to achieve some improvements. Persons are more likely to cooperate when they are involved in the process of change and when that process respects their differences in style and pace.

The STRICU process at LDS Hospital has supported people in multiple roles, depending on their readiness for change. The respect for diversity of styles may be best seen in the approach to one crucial part of the STRICU's improvement efforts: the development, testing, and refinement of standardized, evidenced-based protocols for clinical care [20]. A physician can assume the role of energetic leader, active participant, or passive supporter and reviewer or can opt out of the effort entirely. This last option allows physicians to choose not to have their own patients treated by protocol, on the condition that they do not obstruct the efforts of others. All roles are respected; even the opinions of those who do not support the process are found to be of value as instructive criticism and in reevaluating the effort's process and direction. All attending physicians receive copies of the developing protocols with a letter encouraging them to submit their opinions freely. By encouraging feedback and respecting and negotiating diverse viewpoints, the protocol is improved and cooperation is facilitated during the implementation phase.

In the presence of physicians, other professionals do not always express their opinions. Silence does not necessarily mean agreement or lack of a better idea. When meeting with nonphysicians for the purpose of decision making, an individual physician can increase the pool of ideas by asking each person individually for his or her concerns, ideas, and opinions (Table 2).

Learn How To Negotiate Agreement

Students of cooperation have described useful negotiating techniques for achieving otherwise elusive agreements. Strong leaders master the arts of dialogue, group communication, and negotiation, as taught by experts, and they both model and teach these approaches in the work setting.

In the STRICU, formal sessions on communication and interpersonal dynamics have been taught over the years. Many of the staff have completed the Seven Habits educational series [21] and the Human Dynamics Course [22]. Negotiations stay focused on the issues and not on individuals. Objective criteria for outcomes and cost are used whenever possible, and the opinions of respected peers are used to gain acceptance. An honest attempt to find a win-win solution is always made. One of the fundamental negotiated agreements common to all protocols is that the care team is responsible for the performance of the protocol. When the protocol is not performing correctly, the team can override it. They then communicate the reasons why the protocol was overridden. This information is used to improve the protocol and widen its application. With this process, the STRICU is able to obtain consensus for the protocols.

A physician in conflict with others about what change to make to improve a system can help resolve the conflict by offering two or three alternative solutions. These solutions should be medically sound, should meet the shared aim of the group, and should incorporate in various ways the viewpoints of other disciplines (Table 2).

Insist on Fairness and Equity in Applying the Rules

The rules for the tit-for-tat strategy in the Prisoner's Dilemma are simple; defection is punished in the next round. Clearly defined rules are an important attribute of a cooperative environment.

In the STRICU, those who decide not to cooperate in the development of protocols are not "punished" because punishment could hinder acceptance of the protocols in the future. However, those who opt out are allowed to change their minds and join the process later as circumstances change. Other rules for participation are defined, and the rules for participation are applied equitably. Everyone knows the time limit for commenting on the protocols and knows that nonresponse implies consent. They also know that their critique will modify the protocol directly or through negotiation. Fair application of the rules across all disciplines establishes the integrity of the system and strengthens cooperation. If the rules become difficult to follow, they will be renegotiated but will not be waived for individuals.

Physicians, by virtue of their status in the health care system, can further a spirit of fairness. Once a course of action has been established, physicians are in a position to initiate discussion of consequences and defaults (such as "nonresponse implies consent") for all involved, particularly physician-colleagues (Table 2).

Conclusions

Cooperation does not mean that physicians must give up leadership roles. On the contrary, shifting the meaning of leadership should strengthen it. The physician who wishes to help improve the systems of care must understand that he or she is a central participant in a larger whole that, when correctly led, serves the same ends that the medical profession does. For decades-even centuries-the profession of medicine has been tutored toward excellence in understanding what its elements are, what it is made of, and how it gets better at what it does. Now, in a time of both promising and fearful transformation of health care, the profession of medicine must learn equally well to ask how we will get better at what we must accomplish together if we are to accomplish it at all.

Ms. Spuhler: Shock-Trauma Intensive Care Unit, LDS Hospital, 8th Avenue and C Street, Salt Lake City, UT 84143.

Dr. Berwick (Series Editor): Institute for Healthcare Improvement, 135 Francis Street, Boston, MA 02215.

Dr. Nolan (Series Editor): Associates in Process Improvement, 1110 Bonifant Street, Suite 420, Silver Spring, MD 20910.