Fifteen years have passed since the publication of “Who was caring for Mary?” which is the story of the inexorable downhill clinical course of Mary, my former wife, and a then 34-year-old mother of two.1 As an academic physician, I (F.S.) was especially disappointed by the inattention, the misdiagnoses, and the errors of my colleagues that precipitated this crisis. I blamed academic medical centers (AMCs) for creating conditions that encouraged compromised care, and I called on academic leaders to reward clinical (not just research) excellence. A closer look shows that I assigned too much blame to individuals, too little to the management-of-care delivery processes. While Dr. Francis Peabody's2 famous quotation that “the secret of the care of the patient is in caring for the patient” may have been true in the prior century, his statement fails to take into account the complexity of health care delivery in the 21st century. Because of advances in medical science, effective diagnosis and treatment require excellently coordinating dozens—if not hundreds—of individual professional contributions. Unfortunately, too often, AMCs fail to properly orchestrate the many physicians who contribute to a patient's care.3
Fortunately, those who work in and lead AMCs can solve this problem. Outside of health care, organizations such as Toyota, Alcoa, and the U.S. Navy nuclear submarine program successfully create exceptionally high-performing processes despite the complexity and variety of demands that they face.4 First, these organizations are fastidious about specifying in advance how they expect work to proceed, thereby increasing the likelihood of success. Second, their initial fastidiousness in design renders them extremely sensitive to developments that are contrary to their specific protocols, allowing them to quickly recognize, contain, correct, and prevent errors. Subsequently, they investigate and continually modify their procedures to create an ever better system. By defining normal, recognizing abnormal, and responding quickly, these organizations are self-correcting and self-improving.5
Applying this understanding of how high-performing systems function, we have reevaluated Mary's case and discovered that the many deficits in care previously attributed solely to unprofessional individuals were, in fact, caused by system breakdowns with predictable consequences. We summarize and discuss the facts of Mary's case below to illustrate these points.
Systems Analysis of Mary's Case
In retrospect, Mary's mononeuritis in combination with marked eosinophilia, ecchymotic skin lesions, and multiorgan dysfunction (for full clinical details, please see the original article1) strongly suggest a vasculitic neuropathy induced by an allergic reaction to penicillin.6 A sural nerve biopsy performed after the initiation of high-dose corticosteroids was nondiagnostic. However, her rapid improvement following the administration of corticosteroids indicated that her disease was steroid responsive, and in all likelihood, earlier initial administration of corticosteroids would have prevented many, if not all, of her subsequent complications. With these assumptions in mind, we reexamine each stage of her illness in terms of system breakdowns (as well as individual shortcomings), and we discuss systems designs that could have prevented these breakdowns.
Ironically, the precipitating event was mine (that of F.S.). Mary had pharyngitis with a white exhudate, and our children had recently been treated for culture-proven strep throat; therefore, I prescribed penicillin. In the absence of the rigorous history and work-up that should be performed for all patients including family members, I should not have treated my wife. Mary had no history of penicillin allergy, but unbeknownst to me, a strong family history of allergic reactions to penicillin did exist. Mary should have received care in a facility where appropriate diagnostic tools, including both a thorough history and a throat culture, were available.
Seven days later, Mary developed severe neuritic pain on the bottom of her right foot. She and I (F.S.) sought the help of a neurologist who worked in the university hospital. On the basis of nerve conduction studies, he diagnosed a traumatic nerve injury. Mary denied trauma and an MRI of the right leg was normal. Before the diagnosis and these conflicting findings could be reconciled, the neurologist left for a research conference, and he could not be reached afterwards for follow-up. Perhaps because he was overscheduled—probably a normal situation—the neurologist made several errors in his evaluation of Mary: He ignored the normal MRI, he did not elicit the fact that Mary received penicillin or that penicillin allergies were present within her family, he did not order a complete blood count, and he did not develop a differential diagnosis. Furthermore, on his departure, there was no reliable mechanism for transferring responsibility to a colleague—also a normal condition in many AMCs—though one that threatens continuity of care. Had a handoff actually occurred, another neurologist would have been able to supervise Mary's admission to the hospital, where he or she would have maintained focus on Mary's primary complaint and would have had the highest likelihood of making the appropriate diagnosis.6
On the ninth day of neuritic pain, Mary developed painless bluish-black skin lesions accompanied by ankle swelling. Unable to reach the neurologist, I tried to construct my own transfer of responsibility for Mary's care by contacting an internist in our department whom I highly respected. This physician ordered a venogram that demonstrated venous thrombosis of the right lower extremity, and she recommended hospital admission for intravenous anticoagulation. This internist focused primarily on Mary's thrombophlebitis, did not consider the etiologic significance of the neuritis, did not speak with the neurologist, and did not review or test for the illnesses that can cause neuritis and phlebitis. Continuity of care was compromised again when this doctor had to leave because she was late picking up her children. Because she, too, lacked a reliable way to conduct patient handoffs and a designated person to whom responsibility for patient care should be transferred, the internist rushed her report to another doctor by phone, leaving Mary to yet a third physician who was unfamiliar with Mary's illness and its history/timeline, who failed to focus on the skin lesions, and who failed to perform a skin biopsy. Had the clinic had an appointed long-call physician, charged with person-to-person sign-outs and admissions, this person may have ordered the appropriate tests and consulted with the appropriate specialists—in this case an immunologist familiar with the clinical manifestations of vasculitis. Mary's case would likely not have worsened.
Admission studies showed marked eosinophilia suggesting an allergic reaction. Heparin therapy was initiated but failed to reach therapeutic levels, and on the fifth hospital day Mary suffered a pulmonary embolus. Despite wide availability, the team failed to use anticoagulation order sheets—perhaps because of the absence of an official anticoagulation protocol. As a consequence, the senior team resident and novice intern did not have a clear picture of actions taken and intended effect, so they failed to appreciate that Mary's anticoagulation was not progressing as expected. The senior resident failed to supervise the rotating intern closely and failed to report the underdosing of the heparin to the attending. Both the absence of current laboratory values in the chart and poor communication within the rounding team prevented this experienced physician from intervening. Had an anticoagulation protocol that outlined specific times for partial-thromboplastin-time (PTT) testing and prescribed specific changes in rate of heparin infusion contingent on the PTT results been available, and had specific descriptions of the reporting responsibilities of each team member been in place, Mary's condition would not have further deteriorated.
Furthermore, at this stage, another potentially compromising condition common to AMCs was present. The attending was a division chief with concurrent administrative and research responsibilities distracting him from his clinical duties (much as the neurologist who tried to balance clinical care and travel to a conference may have been distracted). Without these additional administrative and research pressures, the attending may have been better able to focus on Mary's symptoms and clinical needs. Faculty should be assigned as ward attendings only if they can be sufficiently free of other major work responsibilities.
Fever and eosinophilia persisted, and on the eighth hospital day, Mary suffered a myocardial infarction, further indication of a multisystem disease. However, pulmonary, hematology, and neurology consultations proved ineffective. These specialists were thinking and acting only within specialty-specific silos. Not interacting with colleagues blunted the potential benefit of integrating their broad range of knowledge and expertise. The specialists offered various explanations based on their specific proficiencies (vertical care), but Mary's illness was not fully addressed. For instance, no one made any changes in her management despite her continued downward spiral. The consulting specialists should have met as a group to discuss their findings and to develop a unifying diagnosis and treatment plan (horizontal care).
Neither I nor the ward attending realized that Mary was suffering from flaws in the care delivery system. I blamed him alone for the subtherapeutic heparin dosing and for the inability of the subspecialists he consulted to integrate the multiple manifestations of her systemic vasculitis. I lost all confidence in his ability to care for Mary, and I transferred Mary's care to a trusted cardiologist. In retrospect, I realize that the ward attending cared deeply about Mary and that he, too, was saddened by this series of adverse events and lapses that had improbably aligned like the holes in a stack of Swiss cheese to allow Mary's progressive deterioration.
In the final hospital stage, Mary was on the verge of death. She had developed acute respiratory distress syndrome (ARDS), requiring assisted ventilation. ARDS was followed by renal failure and hypotension requiring vasopressor support with dopamine, levophed, and neosynephrine. Suspecting vasculitis, the cardiologist gave her high-dose corticosteroids. At this point, four specialists (a critical care specialist, a cardiologist, a nephrologist, and an infectious disease specialist) communicated openly and provided minute-by-minute clinical care for Mary. Now her care was horizontal and error free, and she received every possible diagnostic and therapeutic intervention. The system was functioning at its best. Whether by design or divine intervention, Mary was saved. But her salvation, as is typical in poorly designed systems, arrived through last-minute heroics, great monetary and emotional expense, and needless risks that would not have been required had the delivery of care earlier in her illness been well managed.
AMCs face greater systems challenges than many nonacademic institutions: The faculty have administrative, research, and teaching obligations that often conflict with patient care (as in the case of the neurologist and the attending); the complexity of diseases is high, necessitating the coordination of multiple subspecialist consultants (Mary had multiple, complicated symptoms); clinical departments can serve as roadblocks to communication (subspecialists offered their own diagnoses, but they did not discuss Mary's case together); and novice physicians, who require close supervision, are left with primary responsibility for the day-to-day care of acutely ill patients (a rotating intern cared for Mary). These challenges require even more carefully coordinated systems; however, many academic physicians work in systems that have too little built-in reliability, efficiency, and safety. Furthermore, many clinical faculty are unaware of how they can personally contribute to improving these systems. Systems seem abstract, and physicians are interested in concrete examples and real patients. Mary's case illustrates the consequences of flawed systems. Given the complexity of medical care, human error and operational breakdowns will always occur,7 but medical care must be carefully choreographed to decrease the chances of these errors and lapses. Increasing the chances for successful execution of care processes requires that physicians embrace the experiential learning cycle of “experience-observe/reflect-conceptualize-retry” that is the backbone of never-ending refinement of complex processes.
Interestingly, the institution where Mary received her care has fully embraced these principles. The institution established a chief of “health care quality and patient safety” position and initiated programs to improve organization accountability, to improve coordination and transitions of care, and to reduce variations in practice. In fact, the institution where Mary received her care is now one of the leaders in health care systems and quality improvement, and the probability of Mary's case now recurring is extremely low.
A large number of innovative systems and quality improvements are developing that academic physicians need to embrace and pass on to their trainees. Web sites that describe playbooks and communication protocols for inpatient teaching rounds8 are a mouse click away, as are Internet sites that allow interactive sessions that demonstrate the fundamentals of patient safety, the principles of teamwork, the methods of improving care systems and measuring their impact, and the processes for effecting change.9
As the reanalysis of Mary's case illustrates, families and patients expect that physicians and the system in which they work will function as a single entity to affect better patient outcomes. Improvements in system performance and professional development must be tightly coupled.10 However, if systems of care are flawed, patients and families will perceive even proficient and compassionate academic physicians as inept and uncaring. We academic physicians must devote our energies not only to improving our personal skills, but also to constantly improving our delivery systems, if we are going to provide the quality of medical care we all aspired to when we entered our profession.
The authors would like to thank Dr. Cynthia Mulrow and Dr. Robert Wachter for their helpful suggestions. They particularly thank Dr. Frank Davidoff for his many insights and invaluable editing. They would also like to thank the Harvard Macy Institute for providing a forum for exploring the implications of introducing systems management practices from high-performing organizations into health care.