History of Health Care Quality

Efforts to assess quality in health care extend back to Dr. Ernest A. Codman, an early twentieth-century surgeon at Massachusetts General Hospital in Boston. Codman was the first to advocate for the tracking and public reporting of “end results” of surgical procedures and initiated the first morbidity and mortality conferences.¹ Such public information would allow patients to choose among surgeons and surgeons to learn from better performers. Codman was clearly ahead of his time. The medical establishment resisted having their “results” measured and publicized, and Codman was accordingly ostracized by his peers. In 1913, however, the American College of Surgeons adopted Codman’s proposal of an “end result system of hospital standardization” and went on to develop the Minimum Standard for Hospitals. These efforts led to the formation of the forerunner of today’s Joint Commission (JC) in the 1950s. The JC’s accreditation process, local hospital quality reviews, professional boards, and other systems that developed allowed the medical profession and hospitals to judge the quality of their work and be held accountable only to themselves for most of the twentieth century.

The academic science of quality management in health care is credited to Dr. Avedis Donabedian’s efforts in the 1980s. Donabedian advocated for evaluating health care quality through assessment of structure (e.g., physical plant, personnel, policies and procedures), process (how things are done), and outcome (final results).² His ideas were first adopted in public health and later spread to administration and management. Shortly thereafter, Dr. Donald Berwick, a Harvard Medical School (Boston) pediatrician, building on a systems approach to quality management, introduced the theory of continuous quality improvement into the medical literature. In his landmark New England Journal of Medicine article,³ Berwick remarked the that the “Theory of Bad Apples” relies on inspection to improve quality (i.e., find and remove the bad apples from the lot). In health care, those who subscribe to this theory seek outliers (deficient health care workers who need to be punished) and advocate a blame and shame culture through reprimand in settings such as morbidity and mortality conferences. The “Theory of Continuous Improvement,” however, focuses on the average worker, not the outlier, and on systems problems, rather than an individual’s failure.

Examples of bad systems abound in medicine. One example provided by Berwick involves the reported deaths resulting from an inadvertent mix-up of racemic epinephrine and vitamin E.⁴ Newborns in a nursery received the epinephrine instead of the vitamin in their nasogastric tubes. If presented as the mix-up of a benign medication for a potentially toxic one, it is viewed as appalling, and blaming individual negligent behavior is easy. Yet when one notes that the two bottles were nearly identical, one can understand why the system “is perfectly designed to kill babies by ensuring a specific—low but inevitable—rate of mixups.” The Theory of Continuous Improvement suggests that quality can be improved by improving the knowledge, practices, and engagement of the average worker and by improving the systems environment in which they work. The “immense, irresistible quantitative power derived from shifting the entire curve of production upward even slightly, as compared with a focus on trimming the tails” is what makes a systems focus so attractive (Fig. 208.1).

Fig. 208.1 Inspection versus continuous quality improvement (CQI).

Although Berwick and others were publishing new studies critiquing health care’s approach to quality and safety throughout the 1980s and 1990s, not until the publication of two reports by the Institute of Medicine (IOM) did quality and safety capture the public’s attention: To Err is Human: Building a Safer Health System,⁵ in 1999, and Crossing the Quality Chasm: A New Health System for the 21st Century,⁶ in 2001.

To Err is Human focused the attention of the U.S. public on patient safety and medical errors in health care. The report’s most famous sound bite, that medical errors result in the deaths of two jumbo jetliners full of patients in U.S. hospitals each day, gained traction with lawmakers, employers, and patient advocacy groups. This estimate, that 44,000 to 98,000 deaths occur per year in the United States as a result of medical error, would make hospital-based errors the eighth leading cause of death in the United States, ahead of breast cancer, acquired immunodeficiency syndrome, and motor vehicle crashes. These statistics were derived from two large retrospective studies. In the Harvard Medical Practice Study,⁷ nurses and physicians reviewed more than 30,000 hospital records and found that adverse events occurred in 3.7% of hospitalizations. Of these adverse events, 13.6% were fatal, and 27.6% were caused by negligence. The Colorado-Utah study⁸ reviewed 15,000 hospital records and found that adverse events occurred in 2.9% of hospitalizations. Negligent adverse events accounted for 27.4% of total adverse events in Utah and for 32.6% of those in Colorado. Of these negligent adverse events, 8.8% were fatal. Extrapolation of results from these two studies provided the upper and lower limits for deaths associated with medical errors for the IOM report. In both studies, compared with other areas of the hospital, the emergency department (ED) had the highest proportion of adverse events resulting from negligence.

Crossing the Quality Chasm focused more broadly on redesign of the health care delivery system to improve the quality of care. The report begins by stating: “The American health care delivery system is in need of fundamental change. … Between the health care we have and the care we could have lies not just a gap, but a chasm.” The report goes on to detail that this chasm exists for preventive, acute, and chronic care and reflects overuse (provision of services when potential for harm exceeds potential for benefit), underuse (failure to provide services when potential for harm exceeds potential for benefit), and misuse (provision of appropriate services, complicated by a preventable error, such that the patient does not receive full benefit). The report first defined six domains for classifying quality improvement for the health care system (Table 208.1). Together, the IOM reports solidified a key concept—quality problems are generally not the result of bad apples but of bad systems. In fact, well-intentioned, hard-working people are routinely defeated by bad systems, regardless of training, competence, and vigilance.

Table 208.1 Institute of Medicine’s Six Aims for Quality Improvement

Adapted from Committee on Quality Health Care in America. Crossing the quality chasm: a new health system for the 21st century. Washington, DC: National Academies Press; 2001.

Costs in Health Care

In 2009, the United States spent $2.5 trillion, or $8000 per person, on health care.⁹ Although we spend well over twice as much per capita on health care than any other industrialized country, we repeatedly fare poorly when compared with other systems on quality and outcomes.¹⁰ Half of the U.S. population admits to being very worried about paying for health care or health care insurance.¹¹ One in four persons in the United States reports that his or her family has had problems paying for medical care during the past year.¹² Close to half of all bankruptcy filings are partly the result of medical expenses.¹³ U.S. businesses similarly feel shackled by health care costs, with reports of companies spending more on health care than on supplies for their main products. In Massachusetts, annual health care costs in school budgets outpaced state aid for schools and forced schools to make spending cuts in books and teacher training.¹⁴ The increased cost of U.S. health care threatens the ability of our society to pay for other needed and wanted services.

Variability in Health Care

Although the number of errors coupled with increasing health care costs paints a negative picture, the variability of care is both more concerning and a source for optimism. Dr. John (Jack) E. Wennberg’s work in regional variability has illustrated the differences in cost, quality, and outcomes of U.S. health care since the 1970s. In 1973, Wennberg and Gittelsohn¹⁵ first identified the extreme degree of variability that exists in clinical practice by documenting that 70% of women in one Maine county had hysterectomies by the age of 70 years, as compared with 20% in a nearby county with similar demographics. Not surprisingly, the rate of hysterectomy was directly proportional to the physical proximity of a gynecologist. Similarly, residents of New Haven, Connecticut are twice as likely to undergo coronary artery bypass grafting (CABG), whereas patients in Boston are twice as likely to undergo carotid endarterectomy, even though both cities are the home of top-rated academic medical centers.¹⁶

Dr. Wennberg’s work in regional variability led to the creation in the 1990s of the Dartmouth Atlas Project, which maps health care use and outcomes for every geographic region in the United States, and details care down to the referral region of each hospital. Atlas researchers, notably Dr. Elliott Fisher et al.,^17,18 have looked at the pattern of health care delivery and have found no clear association between the volume, intensity, or cost of care and patient outcomes. According to the Dartmouth Atlas Project, patients in areas of higher spending receive 60% more care, but the quality of care in those regions is no better, and at times worse, when key quality measures are compared.^17,18 More widely reported in the lay press was the finding that Medicare spends twice the national average per enrollee in McAllen, Texas than it does any other part of the country, including other cities in Texas, without better-quality outcomes.¹⁹ The increased expense comes from more testing, more hospitalizations, more surgical procedures, and more home care. Variability in practice pattern, however, is not limited to overuse, but also includes underuse. McGlynn et al.²⁰ found that, in aggregate, the U.S. population receives only 55% of recommended treatments, regardless of whether preventive, acute, or chronic care is examined.

The findings of the Dartmouth Atlas Project have frequently been cited as a rationale for health care reform law. The Project’s underlying methodology, however, has been scrutinized. Some investigators have questioned whether it appropriately adjusts for the costs of medical practice in different regions. Others have pointed out the limitations of focusing analysis on health care costs incurred by patients over the 2 years before their deaths and attributing these costs to the hospital most frequently visited.²¹ A study by Romley et al.²² reported an inverse relationship between hospital spending and inpatient mortality. Although great variability clearly exists in clinical practice, what is not clear is how to translate maps of geographic variation into health care policies that improve quality and contain costs.

One attempt at decreasing practice variability has been the rapid proliferation of clinical practice guidelines. Guidelines have been developed by physician specialty societies, hospitals, employers’ consortia, and government agencies (e.g., the Agency for Healthcare Research and Quality). The physician community, however, has been slow to adopt guidelines for several reasons. Some physicians disparage guidelines as “cookbook medicine,” whereas others point to the limited evidence on which these guidelines are built. Fear that deviation from guidelines may lead to more malpractice claims has also dampened enthusiasm for their use. Although clinical practice guidelines, order sets, and electronic decision support may not individually be the magic bullet for decreasing unwarranted clinical practice variability, all health care efforts should aim to be safe, effective, patient centered, timely, efficient, and equitable.

Institute of Medicine Aim: Patient Safety

Echoing the ancient axiom of medicine, Primum non nocere, or “First, do no harm,” patients expect not to be harmed by the very care that is intended to help them. As such, patient safety is the most fundamental of the IOM’s six domains. Tables 208.2 and 208.3 provide the IOM’s Patient Safety and Adverse Event Nomenclature.

Table 208.2 Institute of Medicine’s Patient Safety and Adverse Event Nomenclature

Adapted from Kohn LT, Corrigan J, Donaldson MS, McKenzie D. To err is human: building a safer health system. Washington, DC: National Academies Press; 2000.

Table 208.3 Active and Latent Failure Types

Active and Latent Failure Types

Failures in the ED fall into two main failure types: active and latent. Active failures are unsafe acts or omissions at the level of the front-line operator, such as the emergency physician (EP), nurse, or care provider, and the effects are felt almost immediately. This is sometimes called the sharp end of care. Latent failures are failures of the system that can lie dormant, or latent, for years. Despite, or perhaps because of, their obscurity, latent failures can cause multiple types of operator, or active, failures, yet go unnoticed to the casual observer; thus, they pose the greatest threat to safety in the complex ED system. Masked as the cause of incidents, these failures are powerful in joining with other factors to breach the system’s defences and coalesce in errors. In large part, latent failures are the result of decisions affecting daily ED operations, decisions often made by persons not directly involved in care delivery; for example, managers, designers, procedure writers, and drug manufacturers.

Active and latent failures are plentiful in the ED. An active failure occurs when a triage nurse accidentally measures a pediatric patient’s weight in pounds but writes down the weight in a chart box labeled kilograms instead. A similar error occurs when an EP orders antibiotics for a pediatric patient based on milligrams per kilograms (mg/kg) but uses weight in pounds instead. A latent failure exists when a patient with a pending pregnancy test undergoes computed tomography imaging before the positive result returns, or a patient with a pending creatinine test is given intravenous contrast before the abnormal results are acknowledged by the ordering provider.

Reason’s “Swiss Cheese” Model

Described by James Reason in 1990,²³ the “Swiss cheese” model of human error has four levels of human failure useful to evaluating a medical error (Fig. 208.2). The first level depicts active failures, which are the unsafe acts of the operator that ultimately led to the error. The next three levels depict latent failures (preconditions for unsafe acts, unsafe supervision, and organizational influence), which are underlying holes (or hazards) that allow errors to pass through to the sharp end. As with slices of Swiss cheese, when the levels of error are not aligned, an active failure can be caught before it causes harm—a near miss. With the right alignment, however, patients can be harmed by predictable human errors when systems are not appropriately designed to protect them.

Fig. 208.2 Reason’s “Swiss cheese” model.

(Adapted from Reason J. Human error. Cambridge: Cambridge University Press; 1990.)

Cognitive Errors

Cognitive errors fall into three distinct types: skill based, rule based, and knowledge based. Consider the act of driving a car or putting on your shoes. At some point, most of us acquired the requisite skills to accomplish these tasks quickly and efficiently without having to think about them. Skill-based cognitive performance refers to such acts. In the clinical setting, experienced clinicians approach the tasks of preparing a wound, tying a suture, or starting a central line in like fashion. Such clinical actions require little cognitive decision making. Skill-based errors are known as slips and lapses. Slips arise when actions fail to proceed as planned; for example, when a physician chooses an appropriate medication and writes 10 mg when the intention was to write 1 mg. Slips are errors in execution resulting from a failure of attention or perception often caused by interruptions or altered routines. In lay terms, slips are often equated with minor incidents. In the ED, patients can die as a result of slips. Lapses also result in failure to execute a plan, but whereas slips are observable, memory-based lapses are not. Pressing the wrong button on a defibrillator because you are interrupted and lose your train of activity is a slip; not being able to recall the correct energy to defibrillate ventricular stimulation is a lapse.

Any task departure from skills-based processing requires either a rules-based or a knowledge-based approach. Rules-based processing occurs when the clinician applies a known rule to make a decision. Rules, typically applied in the form “if X, then Y,” come from past experience, explicit instructions, or clinical guidelines. Traditional medical education is full of experienced-based rules, often termed clinical pearls. For example, the advice not to discharge a patient with abnormal vital signs is based on clinicians’ cumulative experience of reviewing adverse events after ED discharges. In contrast, the use of formalized clinical decision rules or guidelines (e.g., the Ottawa ankle rules to determine need for radiographs in ankle injuries) is another form of rule-based processing. Knowledge-based processing is when medical knowledge is applied and analytic processes are used to execute a plan of care. Errors of rule-based and knowledge-based cognition are known as mistakes. In rules-based errors, the wrong rule is selected, applied, or linked to the situation. Knowledge-based errors result when incomplete or incorrect knowledge is applied or flawed analytic processes are used, resulting in a poor plan of care. A mistake in medicine may involve selecting the wrong drug or treatment because of an incorrect diagnosis.

As clinicians gain experience, they engage to greater extent in skill-based and rule-based processing and to a lesser degree in knowledge-based processing. This shift in cognitive processing creates an interesting paradox with respect to the types of errors clinicians are most likely to commit. Although the rate of their knowledge-based errors is substantially reduced, highly trained individuals are more likely to experience skill-based errors, which are errors that arise from processes requiring the least amount of cognitive function.

EP are at particular risk for cognitive errors related to the core tasks of the specialty of emergency medicine—the rapid evaluation of patients with undifferentiated complaints and the high-risk decisions that must be made with incomplete information.

Heuristics

Medical decision making in the ED is also characterized by a reliance on heuristics. Heuristics are shortcuts, rules of thumb, or any kind of abbreviated thinking that accomplishes quick and efficient decision making. Although they often serve EPs well, heuristics sometimes fail, thus leading to poor outcomes.

The four most commonly applied heuristics in emergency medicine are representativeness, availability, anchoring, and premature closure. The representativeness heuristic is applied when a clinician makes a subjective judgment of the similarity of a particular patient’s presentation to that of most patients who present with a particular condition. The more unrepresentative the patient’s presentation is, the greater is the chance that the diagnosis will be delayed or missed. Representativeness errors appear most often in settings of high diagnostic uncertainty, such as that typifying the ED, and they are more likely to be committed by clinicians with lower levels of experience. An example is the evaluation of a patient for thoracic aortic dissection, a relatively rare diagnosis. Although the classic symptom of a sudden onset of tearing thoracic chest pain that radiates to the back is not present in most cases, most physicians are less likely to consider this diagnosis without a representative presentation.

Another heuristic that can lead to errors in decision making is availability. Certain encounters are more prevalent in our memories, perhaps because they have occurred more recently, but more often because they are emotionally salient. When making diagnoses, human nature leads to an overreliance on encounters that are vivid and to place less importance on those least salient. For example, if a physician has a particularly vivid experience of missing an acute myocardial infarction (AMI) in a young person, the physician may become overcautious in managing similar patients with chest pain. Availability may similarly be increased by indirect experience: a recent discussion with a colleague, a case presentation at rounds, or an article reviewing a particular case. In contrast, availability is decreased by long intervals since encountering a particular disease, or never having previously seen it.

Anchoring results when physicians commit early to a diagnosis and give it undue weight when considering the available data. One way of avoiding anchoring is to ask “What else could this be?” and always to be disciplined in thinking about the differential diagnosis. The tendency to look for evidence that bolsters an original hypothesis is referred to as confirmation bias. Instead of ignoring conflicting data, EPs must look for disconfirming evidence that rejects the initial contention. If anchoring occurs early in a presentation and EPs operate under a strong confirmation bias, they are sure to miss diagnoses. For example, an older patient who presents with epigastric pain and a bulge in the abdomen could have either an incarcerated hernia or an AMI. A clinician who anchors on the diagnosis of incarcerated hernia can easily explain away tachycardia as the result of pain, and subtle electrocardiographic changes as nonspecific, rather than identifying them as the pattern of an early AMI.

Finally, premature closure occurs when a physician makes a quick diagnosis (often based on pattern recognition and not confirmed by appropriate testing), then stops collecting data, and fails to consider other possible diagnoses. Premature closure can occur in any case, but it is especially common when EPs take care of patients who seem to have an exacerbation of a known disorder. For example, a patient with a history of migraine is assumed to have another migraine, rather than an acute presentation of subarachnoid hemorrhage. Another example is a patient with chest pain and ST-segment elevation on the electrocardiogram who is assumed to be suffering from an acute coronary syndrome, rather than thoracic aortic dissection leading to coronary artery dissection.

Awareness of such cognitive biases is crucial, and simply knowing what they look like can help in overcoming them. Avoiding reflexive thinking and taking the time to think about how we think can help to minimize or avoid errors. Decision support tools, such as templated charts with potential diagnoses, or more sophisticated electronic tools that are queued by patient complaints and history, can help physicians overcome some of these biases.

Error-Producing Conditions in the Emergency Department

The ED ranks among the top three hospital locations with the highest risk of error, along with the intensive care units and operating rooms. This finding is not surprising when one considers the combinations of error-producing conditions that exist in the typical ED. Many different providers (e.g., physicians, nurses, technicians) work closely together to care simultaneously for multiple patients of varying medical acuity who present with any imaginable chief complaint, from the most life-threatening to the most poorly defined, with whom these providers typically have no prior relationship. Providers are bombarded with multiple interruptions, and they struggle at times with understaffing and overcrowding while working through disruptive sleep cycles. As such, error-producing conditions in the ED include diagnostic uncertainty, low signal-to-noise ratio (i.e., the incidence of a serious condition or diagnosis [e.g., cauda equine syndrome] is low compared with more common and more benign diagnoses [e.g., musculoskeletal back pain]), high cognitive load, and poor feedback. Lack of continuity of care and unreliable and untimely feedback of patient diagnoses and outcomes make it difficult for EPs to refine their clinical skills continuously. A national survey of U.S. EDs documented commonly reported problems in four major areas: physical environment, staffing, inpatient coordination, and information coordination and consultation. The surveys suggest that the U.S. EDs have substantial room to reduce the latent conditions for errors.

High-Reliability Organizations

In the ED, we can learn from other industries (e.g., aviation, nuclear power) that have developed and incorporated systems allowing for early error corrections or prevention and leading them to function as high-reliability organizations.²⁴ Such industries tightly couple the process of doing work with the process of learning to do it better. Operations are expressly designed to reveal problems as they occur. When problems arise, no matter how trivial, they are addressed quickly. High-reliability organizations share the following six traits:

Understanding the importance of a culture of safety in the strive for high quality, Johns Hopkins Hospital (Baltimore) researchers implemented a comprehensive, unit-based safety program (CUSPS) at almost 150 individual units in the hospital.²⁵ This program ultimately led to the establishment of an organization-wide culture of safety at the hospital. Although in the past, organizations focused many resources on building a culture of safety, patient safety leaders have shifted to advocating for a just culture, one that balances safety with accountability.²⁶ Whereas a culture of safety allows evaluation of suboptimal medical outcomes without fear of punitive action toward individuals, a culture of accountability also encourages providers to do the right thing, especially when doing the right thing is easy. A notable example is facilitating hand hygiene through provision of soap dispensers in key locations while holding providers accountable for their individual behavior through direct observation and feedback.