Section 4: Patient Safety and Quality Improvement

INTRODUCTION

In the last 15 years, Quality Improvement and Patient Safety have emerged as major focus areas for health care systems around the world. The landmark 1999 report, To Err is Human, defined Patient Safety as freedom from accidental medical injury, which is often the result of error. Errors are defined as failures of execution or planning. Unplanned events that arise from medical care, whether due to human or systems-based errors, are further classified into near-misses or adverse medical events; a near-miss (or “close call”) is an event that causes no harm but had the potential to do so, while an adverse medical event causes patient harm.

Identifying adverse medical events as a source of human suffering, the World Health Organization in 2002 recognized that the need to improve Patient Safety as a fundamental principle of all health systems. The concept of Patient Safety offers a positive spin on the more emotionally laden concept of medical error. Traditionally regarded as the result of incompetence, poor preparation or lack of motivation, medical error is now understood as a product of poorly designed systems of care that contribute to harm. The modern view of medical error is that Patient Safety can be produced only in organizations that take a systems-based approach to the problem, recognizing the inherent limits of human performance and the need to engineer the care delivery process in a way that is based on scientific principles. Nowhere is this issue more pressing than in the acute care hospital.

DEFINING THE PROBLEM

Patient Safety emerged as a public health problem following the November 1999 release of To Err Is Human by the Institute of Medicine (IOM). This report described the epidemic of medical errors in the United States, accounting for as many as 98,000 unnecessary deaths per year. The IOM report described an approach to understanding this problem that relied on developments in human factors engineering and cognitive psychology. By focusing on methods to diagnose and improve systems of care, the report pointed to a novel approach for addressing this epidemic.

The IOM report provoked a broad response. The President of the United States directed the Federal health care agencies to review and implement the recommendations outlined in the report. The predecessor of the Agency for Healthcare Research and Quality (AHRQ) issued $50 million in research grants. Accreditation agencies such as The Joint Commission and the Accreditation Council for Graduate Medical Education developed standards and goals related to Quality Improvement and Patient Safety that are required of hospitals as well as residency and fellowship programs. A group of Fortune 500 companies organized themselves into a consortium called the Leapfrog Group in order to encourage large businesses to purchase health care from organizations that met high standards for Patient Safety. Advocacy groups such as the Institute for Healthcare Improvement and the National Patient Safety Foundation created campaigns and collaborative partnerships to spread Patient Safety-related improvements. Local, regional, state, and national government organizations banded together to develop and cooperate on initiatives to reduce medical errors. In short, the To Err Is Human report helped to crystallize a movement in the United States and abroad that brought a new intensity of purpose to enhancing Patient Safety and reducing medical errors.

HOSPITALIZED PATIENTS

Much of the early work on Patient Safety focused on hospitalized patients. This occurred for several reasons. Inpatients were judged to be particularly vulnerable by virtue of their acute illness, comorbidities, and the intensity of the interventions delivered. Hospitalized patients were more accessible to investigators for study. Improvements that affected the system of care were more readily developed and deployed in the hospital compared to settings such as ambulatory care, with fewer centralized resources to support measurement and improvement initiatives. While Patient Safety in ambulatory and chronic care settings is an area of increasing importance, inpatient acute care services remain the mainstay of safety study and intervention.

Given the central role of the acute care hospital in efforts to study and improve Patient Safety, hospitalists are particularly well positioned to serve as Patient Safety champions in their organizations. Hospitalists are directly responsible for health care delivery, giving them firsthand knowledge about how errors and injuries occur. They understand how current systems may play a role in contributing to harm. Most importantly, they are likely to have an informed perspective about the types and methods of improvement that are both feasible and effective. Frontline provider involvement is a critical component of successful improvement projects, and hospitalists are well equipped to participate in a meaningful way.

SCOPE OF THE PROBLEM

EPIDEMIOLOGY OF MEDICAL ERROR

Medical error was long regarded as a rare phenomenon. In the 1980s and 1990s, however, sentinel cases brought widespread attention to this problem. Among the most widely publicized was the case of Libby Zion, a young woman who died at a New York Hospital in 1984 after she was prescribed meperidine and a monoamine oxidase inhibitor—a fatal combination. Ten years later, Betsy Lehman, a young mother and Boston Globe reporter, died of an accidental chemotherapy overdose due to an ambiguous medication order.

A series of subsequent studies found that errors were common, especially among patients admitted to the hospital through the emergency department. The first large, epidemiologic study of medical errors was reported in the New England Journal of Medicine in 1991. The Harvard Medical Practice study examined over 30,000 medical records of patients hospitalized in New York State in 1984.

Investigators learned that 3.7% of patients had an adverse event, defined as an injury due to medical care. These patients had serious adverse events, including those that extended the patient’s hospitalization, or resulted in death or disability. These medical injuries resulted from surgical and medical care at similar rates, though the events that occurred on the medical service were more often judged to be preventable. Indeed, about one in four events was found to be the result of negligence: care that fell below community standards of medical care.

The Medical Practice Study was an affront to the concept that medical injuries are rare events. Although critics challenged the results, the findings have proven robust. Replications of the Medical Practice Study in Colorado and Utah (in the United States), Canada, United Kingdom, Australia, Spain, and France all show substantially similar results. Five to 25% of hospitalized patients experience an adverse event due to medical care during their hospitalizations, and many are preventable. Researchers extrapolated the Colorado and Utah study results to calculate the 44,000 to 98,000 excess deaths reported in the IOM’s To Err Is Human report.

VULNERABLE PATIENTS

Although all hospitalized patients are at risk of medical errors, certain groups seem to be at particularly increased risk. Both extremes of age are particularly vulnerable, perhaps due to their reduced physiological reserve. An error affecting a sick, elderly person may be more likely to result in injury than in a younger person with fewer comorbidities. The same is true for young children. The need to calculate weight-based medication doses confers on children an increased risk due to medication errors. Other patients at high risk include those undergoing neuro, thoracic, or vascular surgery. These are inherently risky procedures and often performed on individuals with multiple or serious underlying comorbidities. Patients admitted urgently are at higher risk than elective admissions. In addition, the number of interventions a patient experiences increases the opportunities for a mishap. In the Adverse Drug Event Prevention study, Bates and colleagues reported that the highest rates of adverse drug events were among patients in the medical intensive care unit. This was due to the greater number of medications and doses these patients received. Medication-related errors and adverse drug events are an area of special interest to researchers and practitioners, since these events account for the greatest proportion of adverse events among admissions to the medical service. Studies that examined adverse drug events among hospitalized patients identified a consistent list of medications that account for a disproportionate share of serious incidents: anticoagulants, antibiotics, chemotherapy agents, narcotics and sedatives, and insulin.

EMERGING AREAS OF RISK

Adverse drug events have been a particularly fruitful area of work in Patient Safety, resulting in the dissemination of improvements in electronic order-entry systems, pharmacy safe practices, and guidelines for use of high-alert medications. Researchers have been tackling the problem of diagnostic error as well, driven in part by the prominence of missed or delayed cancer diagnoses and miscommunication with outpatient providers. Research has focused on the development of methods to understand lapses in critical processes of care, such as communicating and interpreting critical test results, and ensuring timely completion of referrals. Other thought leaders have focused attention on how doctors think. Can we train clinicians to avoid premature closure of diagnostic options by maintaining a broad differential diagnosis? How can we help them to avoid common mistakes, such as confirmation bias or premature conclusions?

Hospitalists are well suited to study and address the risks associated with handoffs and transitions of care. Hospital medicine practice is rife with opportunities to transition patient care to other hospitalists at the end of the shift or the week, to coordinate care with subspecialists and with colleagues in nursing and pharmacy, and to interact with the referring community practitioner. Research shows that hospital discharge is a particularly vulnerable time for patients, and a time when errors may occur for a variety of reasons. Failure to reconcile mediations at discharge may lead to confusion on the part of patients, reducing outpatient medication adherence rates or increasing the risk of adverse drug events, in turn increasing readmissions. Handoffs to community physicians may fail to occur if the hospital discharge summary is delayed, incomplete, or if there is no system in place to transmit information efficiently. Recommended tests and procedures following discharge are often missed. Promising approaches that hospitalists can utilize to address these problems include standardization of handoffs through the use of templated signout forms, electronic communication with referring providers, and hospitalist-staffed postdischarge clinics designed to support high-risk patients closely after discharge but before their primary care follow-up appointment. Communication and coordination between hospitalists and outpatient providers is increasingly identified as a risk area, and future collaboration between providers in different care settings will be necessary to address it.

As health care moves toward value-based payment and delivery systems, overtreatment and overutilization are increasingly identified as areas contributing to Patient Safety risk. In the decade after To Err is Human, underuse has improved while overutilization has not. Overutilization of inappropriately broad spectrum antibiotics can lead to bacterial resistance or opportunistic infections, as in the case of carbapenem-resistant Enterobateriaceae or Clostridium difficile; hospitalists have an opportunity to become champions of antibiotic stewardship. Medical care can also cause indirect harm: excessive or unnecessary CT scans expose patients to radiation and reveal “incidentalomas”—findings of benign or uncertain significance that would otherwise have gone undetected, but that prompt repeat imaging, laboratory studies, or biopsy. Hospitalists must also be aware of incidentalomas that are missed when transitioning to the outpatient setting, and ultimately can cause delays in diagnosis of malignancy. More diagnoses, more testing, and more treatments increase the potential for patient harm.

The advent of Meaningful Use legislation provided incentives to adopt qualifying electronic health records (EHRs) rapidly, creating new Patient Safety vulnerabilities in many hospital and health systems. While technological innovations like Computerized Physician Order Entry (CPOE) systems can improve efficiency and safe care, Health Information Technology (HIT) may have unintended consequences. After implementing CPOE, one institution found that the adjusted mortality rate in their pediatric population increased from 2.8% to 6.3%. Newly described errors are being recognized, such as the right order being placed on the wrong patient through a CPOE system. Other examples of risk include fragmentation in EHRs, lack of HIT interoperability, and poorly-designed user interfaces. Implementation of a new HIT system is a particularly vulnerable time for patients, and must be monitored closely to mitigate patient harm. During new HIT implementation, hospitalists can have a large impact on patient safety by identifying malfunctioning systems and bringing these glitches to attention.

Creating effective interventions relies on a solid understanding of the nature of error in health care, the methods to assess risk in health care organizations, and the tools that are used to develop Patient Safety improvements. These topics are the focus of the remainder of this chapter.

THE NATURE OF ERROR IN HEALTH CARE

HUMAN BASED

Human error is a complex phenomenon, but one that has come into better focus. Students of human error have argued that both human and systems factors contribute to error. By “human factors” we mean the environmental, work conditions, organizational, and individual characteristics that influence work performance. Experts conceive of human performance in several categories: skills, rule-based actions, and performance that rely on novel problem solving. When skills, knowledge, and rules break down or are misapplied, errors occur. These so-called active failures can further be subdivided into errors of execution and errors of planning.

Skills are stereotyped behaviors that require little conscious thought. When the appropriate skill is chosen but carried out incorrectly, the error is classified as a slip. For example, choosing an antibiotic to treat infection but inadvertently setting the wrong rate on the infusion pump would be an example of a slip. Similar to a slip, a lapse is when the appropriate action is omitted or not carried out, rather than being carried out incorrectly. When facing an unfamiliar scenario where one cannot readily apply an automatic or routine skill, knowledge-based and rules-based cognition takes over. Mistakes can be knowledge- or rules-based, and result when an individual did what they intended, but it did not work as planned. For example, recognizing the infection but choosing to treat with a diuretic would be a mistake.

Everyone is prone to slips, lapses, and mistakes. We can reduce the frequency of these errors through education and training. However, no one is infallible, and therefore, no one is immune from error. In fact, certain conditions can increase the risk of harm. Workers are more likely to make slips and mistakes when they are tired or overworked, bored, distracted, intoxicated, or ill.

Skills, rules and knowledge-based cognitive errors are often unavoidable, and therefore clinicians should not be held accountable for these innocent errors. In contrast, individual actors should be accountable for at-risk behaviors, which occur when an individual is unaware of the hazard associated with a behavior that deviates from the standard, and reckless behaviors, which occur when an individual knowingly acts dangerously. These behaviors are rare and may merit disciplinary action.

SYSTEMS BASED

Although all humans err, the impact of mistakes is more serious for individuals whose decisions and behaviors have a consequential effect on others. Military and commercial aviation, nuclear power, and health care are examples of industries where error can be catastrophic. In these settings, researchers and organizational leaders have begun to focus on the systems in which individuals work in order to design defenses that identify, intercept, and prevent errors before they result in harm.

A system is defined as a set of interdependent processes designed to accomplish a common aim. Certain characteristics of systems can allow or facilitate individuals’ performance of unsafe acts. These characteristics are often called latent conditions or latent factors. Examples of latent factors include poor training, duty schedules that provide little time for sleep, lack of adequate supervision, lack of sufficient supplies, and a culture that discourages cooperation and teamwork. Analysts often focus on latent factors that represent design flaws for a particular process and that, in turn, allow unsafe acts to result in harm.

Consider the case of a physician who failed to follow-up on a radiology report showing a new lung nodule. The unsafe act must be understood in the context of the latent factors that contributed to the error. Was the physician overworked, covering for vacationing colleagues? Was there a consistent approach in place for the practice for follow-up of test results? Did a radiologist attempt to contact the ordering clinician unsuccessfully? Multiple latent factors typically contribute to an accident—and few are apparent until after an accident occurs. When a series of latent failures align, harm can result. This model of organizational failure has been described by British psychologist James Reason as the “Swiss cheese model.”

Errors in medicine can be classified in other ways, into categories such as diagnostic errors, medication errors, and communication and transition errors. It is important to examine the underlying contributions of human and systems factors to each of these categories. These issues will be addressed in subsequent chapters.

EFFECTIVE PATIENT SAFETY PROGRAM DESIGN AND CULTURE

No single model has emerged for an “ideal” Patient Safety program in hospitals. There are, however, several key components of effective programs. Effective programs must have methods to detect errors, analyze events, and implement improvements. In addition, effective programs build a safety culture in the organization that fosters an environment where reporting, analyses, and improvement initiatives can flourish. Buy-in from senior leadership and physician engagement is a particularly important ingredient of the mix. Effective Patient Safety programs are resourced adequately to fulfill their mission. Many hospitals appoint and support Patient Safety officers, risk managers, and data analysts, data managers, and process improvement specialists. These professionals and their activities are usually housed in a Department of Healthcare Quality, although the specific arrangements will vary from organization to organization.

MEASUREMENT AND ANALYSIS

Hospitals use a variety of tools and techniques to measure Patient Safety. Most hospitals have voluntary systems for reporting near-misses and adverse medical events. This approach is in widespread use, in part because of government and accreditation agency requirements. Certain serious events must be reviewed internally and reported to the appropriate external oversight agency. Most hospital pharmacy departments use a similar approach, reporting “interventions” that pharmacists perform when they clarify or correct a clinician’s order. Pharmacy interventions and safety incidents together represent an important source of data about errors and injuries, but these methods are subject to reporting bias. Busy clinicians often do not have time to complete these reports and may be less likely to report their own errors than those performed by colleagues upstream in the care process. In addition, clinicians tend to report very few adverse events compared to nurses or other hospital staff, but those that they do report tend to be more serious. In some organizations, an increasing number of incident reports is interpreted appropriately as a sign that safety is taken seriously by front-line clinicians. Hospitalists should consider reporting adverse events, particularly life-threatening ones, as part of their duty to patient care.

Given the limitations of incident reporting and pharmacy interventions, hospital leaders and Patient Safety researchers have examined a variety of alternative approaches. Direct observation of clinicians at the point of care is a fruitful strategy but requires a tremendous amount of time and effort to maintain. Chart review methods are also well established but potentially resource intensive. Since not all errors result in harm, recent measurement tools have been developed to focus on harm events. The Institute for Healthcare Improvement Global Trigger Tool allows for the identification of adverse events, based on clues seen in the medical record.

Some organizations have access to sophisticated tools that can screen electronic medical records for possible errors or adverse events. These tools examine medication records for events—such as an order for antidote drugs like diphenhydramine, naloxone, and epinephrine—that may signal the presence of an adverse event. AHRQ has developed a set of Patient Safety Indicators (PSIs) that screen administrative records for adverse events based on diagnosis and procedure codes.

Once a critical incident or set of incidents has been identified, health care organizations need to cull the lessons that can be learned from these events. Traditionally, this has been the subject of the Morbidity and Mortality Conference, though the focus of M&Ms has often been on individual performance rather than systems factors that contributed to errors. In contrast, root cause analysis is a systematic and structured approach to identify the latent conditions that contributed to an error. Root cause analyses, performed properly, help organizations to learn about the causes of errors and injuries and, in turn, to develop initiatives that prevent these errors from happening again.

IMPROVEMENT IMPLEMENTATION

It is not enough to measure and analyze adverse events and near misses; effective Patient Safety programs integrate quality improvement techniques and personnel to implement interventions designed to prevent harm. Near-misses are particularly important, as they signal weakness in the care delivery process that have not yet resulted in harm but might do so if left unaddressed.

After an event is analyzed, the processes of care leading to that event must be elucidated in order to design an effective intervention. This step is critical. When the processes are mapped, examined, and understood, appropriate interventions can be designed. Improvement interventions vary from changing one step in a provider’s workflow to redesigning entire care delivery processes. The scale and scope dictates the resources and method of implementation necessary to produce reliable, durable changes. Hospitalists not only have opportunity to review adverse events via root cause analyses, they also can aid quality improvement personnel in understanding frontline provider workflows.

FOSTERING A CULTURE OF SAFETY—LEADERSHIP AND FRONTLINE PROVIDER ENGAGEMENT

Other high-risk industries, like aviation and nuclear power, recognized long ago that organizational leadership shapes culture. Obtaining buy-in from hospital boards and executive leadership for Patient Safety programs is essential to securing the engagement of front line providers.

The Institute for Healthcare Improvement (IHI), in their 5 Million Lives campaign, recommended “Getting Boards on Board,” in an effort to “fully engage the governance leadership in quality and safety.” Hospital boards can drive safe care by using the following approaches:

• “Setting aims: Set a specific aim to reduce harm. Make an explicit, public commitment to measurable quality improvements.

• Getting data and hearing stories: Select and review progress toward safer care… at every board meeting, grounded in transparency.

• Establishing and monitoring system-level measures: Identify a small group of organization-wide ‘roll-up’ measures of Patient Safety that are continually updated and are made transparent to the entire organization and its customers.

• Changing the environment, policies, and culture: Commit to establish and maintain an environment that is respectful, fair, and just.

• Learning, starting with the board: Develop the board’s capability and learn about how ‘best in the world’ boards work with executive and medical staff leaders to reduce harm.

• Establishing executive accountability: Oversee the effective execution of a plan to achieve aims to reduce harm, including executive team accountability for clear quality improvement targets.”

While many institutions strive to establish a nonpunitive environment, as of 2014 up to 50% of employees still worry about negative repercussions from reporting. Emphasizing a nonpunitive environment for adverse event and error analysis can help to establish a culture of safety, where all employees feel comfortable disclosing errors knowing that their disclosure will lead to improvement. The importance of front-line provider engagement is essential, and hospitalists play a critical role.

Hospitalists understand the systems and processes of care within their local environment. They can identify potential safety risks, escalate their concerns, advocate for improvements, and evaluate improvement designs critically. Hospitalists can weigh the feasibility of implementation plans and become local champions for process or cultural changes.

QUALITY IMPROVEMENT

Although Patient Safety is a cornerstone of medical care, the notion of “quality” in health care includes several other important components. The IOM has defined six dimensions by which quality in health care can be evaluated:

• Safe: avoiding injuries to patients from the care that is intended to help them.

• Effective: providing services based on scientific knowledge to all who could benefit, and refraining from providing services to those not likely to benefit.

• Patient-centered: providing care that is respectful of and responsive to individual patient preferences, needs, and values, and ensuring that patient values guide all clinical decisions.

• Timely: reducing waits and sometimes harmful delays for both those who receive and those who give care.

• Efficient: avoiding waste, including waste of equipment, supplies, ideas, and energy.

• Equitable: providing care that does not vary in quality because of personal characteristics such as gender, ethnicity, geographic location, and socioeconomic status.

In this formulation, safety is one of several components of high quality care.

IMPROVEMENT METHODS

How does a health system improve its performance in one or more of these domains? Many organizations rely on the Model for Improvement, a well-described approach used to promote organizational change. The basic steps in the Model for Improvement include the following:

• Setting aims

• Establishing measures

• Selecting changes

• Small tests of change, as in Deming’s plan-do-study-act (PDSA) model

• Implementing changes

• Spreading changes

After setting aims, measures must be established to determine whether improvement results from the changes that have been implemented. There are three fundamental types of measures:

• Outcome measures: evaluate the end result of a given system or process.

• Process measures: evaluate the steps involved in a process.

• Balancing measures: evaluate whether changes in one area result in (unintended) changes elsewhere.

Unlike measurement in research, measurement in quality improvement is used to bring new knowledge into daily practice. It consists of small tests of change, with many sequential tests, and just enough data gathered in each round of testing to assess if change results in improvement. The process is then continued with the implementation and spread of change.

Another improvement method that was developed originally for industry uses a detailed study of production processes to improve efficiency by eliminating various forms of waste. The Toyota Production System is an example of this management philosophy, which has become known generically as the Lean production system. Lean methodologies have made their way into health care, with a focus on improving quality and reducing cost by streamlining the complex care delivery process. Reducing complexity and waste reduces resources and risk. A care process that takes 50 steps and 10 people to accomplish is more expensive and potentially riskier than a 10-step process involving two people. While Lean is not engineered to improve Patient Safety, it may do so indirectly by eliminating inefficiencies that increase the potential for error and harm.

QUALITY IMPROVEMENT AND SAFETY RESEARCH

Quality Improvement and Patient Safety have intersected in recent years, as quality improvement methods have been applied to solve Patient Safety problems. For example, health leaders used the PDSA model to develop medication safety improvements for anticoagulants and other high-alert medications. Rapid-cycle improvements have led to innovation in communication of critical test results and handoff communication. As researchers understood the value of human factors principles and system-based design, Patient Safety leaders have embraced and promulgated “best practice” interventions that rely on these concepts. Key principles include the concepts of standardization and reliability, appropriate redundancy, use of communication, and teamwork tools. Best practice recommendations have been incorporated into recommendations and standards put forth by The Joint Commission and the National Quality Forum. Recognizing that there are limits to human performance, researchers have investigated the use of forcing functions as prompts. Attention has also been focused on the limits of human performance during times of fatigue, with efforts to reduce these effects. Each of these concepts will be expanded upon in subsequent chapters.

CONCLUSION

Recognizing the need to mitigate patient harm from preventable adverse medical events, Patient Safety has become an organizational priority of many health care institutions across the world. The epicenter of work in the area continues in hospital care, given the multiple interventions delivered there and the vulnerability of the patient population. New areas of Patient Safety risk are identified regularly and become the focus of scholarly research and improvement initiatives. Effective Patient Safety programs are built on the understanding that medical errors are the result of human factors and systems issues, obtain executive, board, and frontline provider engagement, and establish a culture of safety by reinforcing a nonpunitive environment. The leaders of these programs design systems that identify errors and learn from them, build interventions to prevent errors and mitigate harm, and disseminate their work as best practices to the greater medical community. Organizations that do these things well create safer, more effective patient care.

SUGGESTED READINGS

INTRODUCTION

“Patient-centered care” is a core principle in health care, and the concept continues to move from innovation to expectation and in some cases, even regulation. But what exactly does it mean to provide patient-centered care, and how does one achieve this, at either the individual or institutional level? This chapter explores the term patient-centered, and current thinking about how to strengthen the partnership between patients and providers in the delivery of care. Throughout, emphasis is given to those innovations most relevant to the hospitalized patient.

DEFINING PATIENT-CENTERED CARE

While the term patient centered now appears commonly in both medical literature and lay media, one may encounter a variety of definitions for this phrase. Perhaps the most “official” definition is the one proposed by the Institute of Medicine (IOM) in the landmark 2001 document Crossing the Quality Chasm. It describes patient-centered care as “…care that is respectful of and responsive to individual patient preferences, needs, and values, and ensuring that patient values guide all clinical decisions.” The IOM goes on to describe patient-centered care as one of the six key “aims for improvement” for quality of care. As such it is presented as an intrinsic value, fundamental and irrefutable, as opposed to a system property with a known association with better outcomes. The concept has been advanced in the form of slogans such as “Nothing about me without me,” “Every patient is the only patient,” “You’re a person before you’re a patient,” and “Human First.” At its core, patient-centered care can be described as treating each patient with the respect and dignity that every human being inherently deserves.

Overlapping terms appear in the both the lay and medical literature, including patient partnering and family-centered care. In this chapter, the term “patient-centered” will be used to encompass the general concept of making care delivery more responsive to the needs and wishes of the individual patient and his or her family. Family can be defined broadly as all the individuals whom the patient wants involved in his or her care regardless of whether they are related biologically, legally, or otherwise. From this definition, it follows that if a patient has any family, patient-centered care must include them.

The construct presumes that care delivery under current models is not adequately patient-centered. The IOM “Chasm” report conceptualizes the health care delivery system as in evolution, from a clinician-centric, poorly coordinated and non–evidence-based model to a patient-centric, integrated system that consistently applies scientifically supported interventions. In “clinician-centric” models, the patient plays a passive role in decision making regarding choice, timing, and settings of care delivery; these are the exclusive domain of the providers, and those same providers decide what information reaches the patient. Stories abound in the lay and medical literature of patients feeling at the mercy of the medical system, unable to exert control over their own care. However, it is hard to find a quantitative assessment of the current state of patient-centered care (or lack thereof) in the US health care system. Some insights can be gleaned from national results of the Hospital Consumer Assessment of Health Care Providers and Systems inpatient survey distributed by the Centers for Medicare and Medicaid Services (CMS). Although scores have been improving, 18% of patients still report “never or sometimes” receiving communication about new medications and their side effects, and 9% answer similarly to questions about the responsiveness of hospital staff (composite scores, July 2012-June 2013 data).

The IOM describes the fully evolved stage of organizational development as characterized by the patient and family being part of the health care team, with full access to information and the ability to exercise as much control over care as desired. What specific actions can institutions and providers take to advance toward this model? These can be divided into three key properties: (1) free flow of information, (2) partnering around individual patient needs, and (3) involving patients in system design.

FREE FLOW OF INFORMATION

Patients are not truly partners in their own care if information, either about themselves or the care they are receiving, is only selectively available. The IOM suggests two “rules for redesign” to achieve transparency:

• Knowledge is shared and information flows freely.

• Transparency is necessary.

The first rule recognizes that patients should have the ability to receive complete and understandable information about their condition, in real time. The second “rule” establishes that patients are entitled to information about the care itself, including the performance of the health care system and its providers, as well as the approach to care and its justification. While these concepts may seem self-evident, the health care system has not traditionally been aligned with these rules. Prior to passage of The Health Insurance Portability and Accountability Act (HIPAA), exchange of medical documentation sometimes required that a patient obtain a subpoena. Today, care still remains far from transparent: published reports suggest the majority of hospitalized patients are unable to easily determine the name of the physician in charge of their care, let alone the details of the care plan. Fortunately, recent innovations are improving information exchange with patients and changing longstanding traditions of care delivery.

OPEN COMMUNICATION OF THE PLAN OF CARE

Typically, the discussion, development, and implementation of a hospitalized patient’s plan of care occur without the patient’s involvement. Communication of the plan of care is a separate responsibility of the physician, occurring most often as an unstructured verbal communication. There is thus no system that guarantees that the patient understands the plan, or has had the chance to ask questions. Approaches that provide structure to these exchanges, and thus more reliable sharing of information, are appearing in the interest of both patient-centered care and patient safety. For example, the Veterans Health Administration introduced “The Daily Plan,” a structured document reviewed with the patient each day of his or her hospitalization. It contains information such as medications, scheduled procedures, and diet, reviewed with the patient each day of his or her hospitalization. The expectation is this kind of intervention could improve provider-patient information exchange in both directions, provide an opportunity for patients to ask questions and share concerns, and identify problems that might cause risks to safety or make the provider’s plan ineffective or infeasible. Reported experience shows that the large majority of patients receiving such a plan perceive a better understanding of their hospitalization, have a better ability to ask questions, and a higher level of comfort with their hospital stay. Similar positive findings have been seen by introducing structured patient involvement with hospital discharge planning. A key element of these new models is some mechanism for “closed loop communication,” meaning there is verification that the communication has been received, understood, and any remaining questions have been answered.

Several hospitals have actually embedded communication with patients into the work model by adapting bedside “rounds” to include the patient and/or family member. In a typical format, the patient and/or family member is oriented to the process of rounds and is given the option to participate. On rounds, the patient/family member is introduced to the members of the team, hears the presentation of the clinical situation and is invited to participate in developing the plan. Teaching, including discussion of the condition and demonstration of physical findings, occurs with the patient’s permission. For patients, such programs have been associated with higher satisfaction, better clinical outcomes, and shorter lengths of stay. Health care workers, in turn, have reported higher satisfaction with work and with the quality of some aspects of teaching in academic medical centers, but have also identified challenges that have inhibited widespread adoption of such processes.

COMMUNICATION AND RESOLUTION IN SETTING OF ADVERSE EVENTS

Flow of information should not stop if care does not go as planned. Patients who experience an adverse event are ethically entitled to receive information about that event, and typically respond favorably to “I’m sorry” in those situations where apology is indeed appropriate. Unfortunately, open communication with patients about medical error has been inhibited by a conventional wisdom, held for decades, that open disclosure and apology would increase risk of litigation. There is a growing body of evidence that this is a false premise, and that open communication following error is not only ethically correct, but is likely to decrease malpractice costs since it provides an opportunity to bring issues to resolution without involving the court system. A growing number of institutions are now implementing a systematized approach to harm events that involves open communication, root cause analysis, and early resolution if the analysis determines that the harm was preventable. Many of these programs are now reporting a coincident improvement in malpractice costs. Such programs are also more patient-centered in that time to resolution is quicker, and nonmonetary elements can be part of the resolution. For example, patients suffering harm often look for an institutional commitment to decrease risk of a similar harm event occurring in the future.

Remaining patient-centered in the context of adverse events requires that an institution establish an unambiguous position on the topic, and communicate that position to the workforce. Mechanisms must be put in place to educate and support clinicians in the process of disclosure and apology, which is often an ongoing event, requiring multiple communications as facts become available. For any given clinician, personal involvement in disclosing an error to a patient will be a rare event, so systems for “just in time” support and training must be available. Many institutions address this by creating a resource group with specific interest and training in best practice for communication, empathy, and apology (where appropriate). Physicians, nurses, social workers, patient safety professionals could all potentially serve in such a role. In the setting of an adverse event, the expert resource can support the clinicians involved and help determine the best timing, setting and participants for communicating the event (see Chapter 20: Preventing and Managing Adverse Patient Events).

ACCESS TO MEDICAL DOCUMENTATION

Medical documentation has traditionally been the purview of the clinicians and not the patient. The passage of HIPAA in 1996 established that patients must be permitted to review and amend their medical records, but access to the record is still largely based on an exception process, the record being provided when there is an active request by the patient, which in practice occurs rarely. On the other hand, surveys show that when given the option to view the clinical record, the large majority of patients will accept. Many institutions have responded with systems that allow patients to directly access elements of clinical documentation electronically. This is often limited to objective content such as problem lists, medication lists, and test results.

Ready access to clinician documentation is now gaining momentum. A formal, multi-institutional trial of “OpenNotes,” a system for electronically sharing outpatient clinical documentation, demonstrated that more than two-thirds of patients reported better understanding of their medical conditions, and felt more in control of their care. At the same time, only 3% of physicians reported spending more time answering patient questions outside of visits. The clinical impact of more open access to documentation is now getting more attention. For example, physicians in the OpenNotes trial report they are more attentive to the accuracy of their documentation, and the approach to documentation of sensitive conditions such as cancer, mental health, and substance abuse. Results of the OpenNotes trial were compelling enough that all three participating institutions have adopted the program as standard practice throughout their ambulatory operations. More research is needed in areas such as the potential for open sharing of clinical documentation to better identify medical error, and/or to improve clinical outcomes.

PATIENT-GENERATED HEALTH DATA

Patient-generated health data (PGHD) include biometric data (eg, home blood pressure readings), medical history, symptoms, and administrative information (eg, the name of a surrogate medical decision maker). They are recorded by patients, as opposed to providers, and the growth of Internet-enabled devices has encouraged a movement to collect more PGHD. Proponents of PGHD hope that it will engage patients more in their care and believe that it could shift some of the burden of data collection from providers during time-limited encounters to patients outside of such encounters, freeing patients and providers to spend their time together in more valuable ways, such as more deeply engaging in the shared decision-making process. However, there are many challenges. Information technology professionals must ensure that PGHD interfaces are user friendly, facilitate high-quality data collection, and maintain appropriate privacy. Simultaneously, it is important to ensure maximal interoperability between different electronic medical record systems. Frontline clinicians must be involved in the design of PGHD systems so that the new streams of data will fit into their busy workflows. And all parties must develop ways of reviewing PGHD, especially prior to including them in patients’ medical records.

PATIENT-REPORTED OUTCOMES

Current standards for assessment of clinical outcomes are based largely on objective measures and rarely include the patient’s perspective. For example, outcome of a joint replacement surgery might be evaluated on the basis of range of motion or occurrence of a complication, without formal capture of what the patient has experienced in terms of pain relief or ability to pursue daily activities. For some time, evaluation of clinical outcomes in the research domain has included such patient-reported outcomes (PROs). Now, attention to PROs is growing as part of routine clinical practice, spurred in part by provisions in the US Patient Protection and Affordable Care Act, which specifies a focus on PROs and launched the Patient-Centered Outcomes Research Institute. At the same time, the increasing presence of electronic formats for collecting and storing survey data is fueling the ability to collect PROs.

PROs might include information about which outcomes matter most to patients as well as actual measures of symptoms, functional status, or quality of life. Collection of such outcomes is especially relevant to conditions involving longitudinal care such as cancer, heart disease, mental illness, or arthritis. There is growing interest in developing standardized instruments for collecting PROs, and formal guidance from the National Quality Forum on valid design of PRO-performance measures has been issued. Some of the best examples of broad implementation of PROs in clinical practice come from outside the United States. The National Health Service in the United Kingdom, for example, collects information on symptoms and functional status for all patients undergoing certain elective surgeries.

ACCESS TO INFORMATION ON CLINICAL PERFORMANCE

There is a slow but undeniable trend toward sharing more information about clinical performance with the lay community. Health care institutions have resisted this trend, on the basis that clinical performance data are too difficult to correctly interpret, cannot be adequately risk adjusted, and/or will perversely impact clinician behavior. Thus, much of the initial effort to make performance data public has been involuntary, driven by regulators, creditors, or insurers, each of which may have its own unique requirements for public reporting of clinical performance. As a result, clinical information available in the public domain can vary dramatically from state to state. At a federal level, the patient can find a growing list of measures of hospital performance disseminated by CMS via their website, http://www.medicare.gov/hospitalcompare/search.html.

Simultaneously, many institutions are now choosing to voluntarily share information on clinical performance (Figure 16-1). This trend is most readily apparent as a component of hospital websites, but some hospitals are also choosing to share information in the form of mailings or posted material within the facility (illustration). Reasons to pursue this strategy likely vary by institution, but might include (1) the ability to provide context and explanation to information already being shared elsewhere, (2) the ability to determine and expand the portfolio of information being shared, (3) the belief that sharing clinical performance information is a good business strategy, (4) the belief that sharing clinical performance information is consistent with institutional values. The result is that a current survey of hospital websites will demonstrate a broad range of approaches to transparency: some provide a great breadth of information on performance, others almost none. Some hospitals provide metrics with a minimal amount of explanatory information, while others appear to go to great lengths to make the information accessible and available to a lay audience.

Few data are available regarding the degree to which patients use clinical performance data to make decisions regarding their own care. The data that do exist suggest that public opinion strongly favors the concept of public sharing of performance data, even though few consumers currently direct their care on the basis of objective, publicly reported metrics.

PARTNERING AROUND INDIVIDUAL PATIENT NEEDS

Assuming the first ideal has been met and there is complete transparency and flow of information; how much control does the patient have over the plan of care, or the ways in which that plan is carried out? Patients often encounter systems that are unresponsive to their individual needs. Truly patient-centered care requires institutions to have the capability to fully elicit individual patient preferences, and the capacity to reliably customize care in response to those preferences.

APPRECIATION OF INDIVIDUAL NEEDS

At one level, the system can be made more responsive to the patient simply by ensuring that the clinicians have a complete understanding of individual preferences. Unfortunately, despite the central role of interpersonal communication in almost every aspect of care delivery, evidence suggests that clinician-patient communication is often imperfect, and communication skills have only relatively recently been recognized as a core competency by entities such as the Accreditation Council for Graduate Medical Education and The National Board of Medical Examiners. Training can in fact lead to improved communication skills, allowing the clinician to better identify the unique needs and values of the patient, while simultaneously improving patient understanding of his or her clinical situation.

Since needs and preferences may be linked to individual ethnic and religious traditions, patient-centered care requires that providers have an appreciation for this context. Cultural competence refers to the provider’s ability to bridge cultural differences in the provider-patient relationship, through understanding and respect of the patient’s beliefs and awareness of one’s own biases. Cultural competency is most often discussed as a strategy for decreasing the persistent inferior clinical outcomes among minority populations (see Chapter 37: Cultural Competency in Healthcare), but it is equally clear that cultural competency is an essential component of patient-centered care. Like communication, cultural competency is teachable; training programs in cultural competency are on the increase, and in fact are mandated for physicians in many states. In addition, institutions with significant numbers of patients from specific ethnic communities should look for ways to partner with those communities to develop both a shared appreciation of specific needs and preferences as well as appropriate institutional supports.

CAPACITY TO PREPARE FOR AND RESPOND TO INDIVIDUAL NEEDS

Full appreciation of the unique needs of the individual is only meaningful if the system can customize care in response to those needs. The concept of partnering with patients is one way to think about the process of customizing care. Helping patients understand the reasons underlying their care is likely to make them engaged participants rather than passive recipients, and therefore more likely to lead to patient-centered care. For instance, imagine a patient who just had a hip replacement and is asking to leave the hospital as soon as possible. One approach would be to tell her she cannot yet leave, and that the health care team will decide when she is ready to be discharged. A partnering approach would start by asking about her concerns (Patient: “I am worried about getting a hospital-acquired infection”), align with those (Hospitalist: “We do not want you to get an infection or stay in the hospital any longer than is necessary”), help her understand the events that need to occur before it is safe for her to leave the hospital (Hospitalist: “In order for you to be able to function when you leave the hospital, we feel it’s important that you be able to…”), and finish by asking if the team has addressed her concerns (Hospitalist: “How does that sound to you? What questions do you have?”), thereby making her a more active participant in her care.

In some cases, customizing care requires some preparation. For example, encouraging patients to legally designate a surrogate medical decision maker is a critical part of advance care planning; without that information, by the time the patient loses capacity to make their own medical decisions, it may be too late to determine what their preferences are and who they would prefer to speak on their behalf. Other examples include reliably recording who the patient wants involved—and who they do not want involved—in their care, as well as how he or she prefers to be addressed, and which language is best for him or her. Customized patient-centered care then means consistently using that information when speaking with the patient and family, as well as arranging any necessary interpreter services in advance of an encounter. Electronic medical record systems must be designed to support busy clinicians by making it easy to record such information, and then ensuring it is easily retrieved at the point of care.

Being responsive to individual needs also requires ceding some control to the patient. For example, many hospitals are eliminating restrictions on visiting hours in favor of open access for the patient’s family. Family presence during invasive procedures and resuscitation events is endorsed by several professional societies. Programs that allow patients or family members to activate rapid response teams are now well described. Sometimes called “Condition H” or “Code Help,” the concept was initially advanced for the pediatric setting, where serious adverse events have occurred in the presence of a concerned parent who was unable to bring immediate assistance to the bedside. The same principal has now been extended to the adult setting. A typical response team will include both a physician and nurse, and in some models may include critical care specialists, social workers, chaplains, and/or patient representatives. It is not unusual for clinicians to initially express resistance to the concept, citing concerns that patients and family members will overuse “code help” for inappropriate, nonurgent issues. However, the experience of institutions that have implemented response teams is that the option is used prudently, and allows earlier interception of potential adverse events as well as other important issues such as inadequate pain control or communication problems. Successful implementation of a “code help” program requires patient and family education about its purpose, a well-structured activation mechanism, and a predefined set of individual responders. Protocols that define the mechanism for recording the incident, and debriefing with the patient and family, should also be established.

Partnering around individual needs becomes particularly important when treatment options carry significant risks or consequences to the patient. Conditions with multiple potential approaches to management such as chronic back pain, depression, cancer, or organ failure, are all examples of conditions where patient preference plays a major role in determining the best plan of care. Such situations call for informed, shared decision making, a term for the process of communicating with the patient about what matters most to him or her in the context of the medical situation, with the ultimate goal of making patient-centered medical decisions. Shared decision making and partnering with patients replace two inappropriate alternatives: telling patients what to do without their input, and asking them to make choices from a list of options without any guidance or advice. Education about shared decision making may help clinicians and patients navigate this complex process and find the right balance.

The first step is to ensure the patient understands his or her condition, the prognosis, and the treatment options, to the degree they desire. This can be accomplished using established communication techniques such as “Ask-Tell-Ask”: (1) “ask” the patient how much they would like to know, how involved they would like to be in decision making, and what they understand; (2) “tell” them about what they do not yet understand in patient-centered language; (3) “ask” them how they feel about that information, what questions they have and to reiterate their understanding to ensure everyone has a shared understanding.

Before discussing treatment options, it is important for clinicians to understand what matters most to the patient, in the context of the medical situation. Although it is a seemingly vague construct, understanding “what matters most” becomes increasingly important as the stakes of the medical decisions rise. Asking about the patient’s hopes, goals, fears, and worries, and about any tradeoffs they are willing to make in order to reach their goals, brings clarity to what matters most to a patient, and provides invaluable information for the next step in shared decision making.

Discussion of treatment options must include information about the anticipated benefits and risks of each option, and should include a recommendation from the clinician based on knowledge of what matters most to the patient. Without such knowledge, the clinician risks suggesting options that will not actually benefit the patient. For instance, recommending an aggressive chemotherapy regimen may be inappropriate for a patient who prefers to maximize quality of life and minimize time spent in health care settings, even if that means they may die sooner than they would if they received that regimen. Informed shared decision making can be supported by decision aids, that is, structured materials such as a videotape or printed algorithm that help to illustrate treatment options and associated risks and benefits.

INVOLVING PATIENTS IN SYSTEM DESIGN

Systems of care delivery have largely evolved in response to the needs of providers and the design of the payment system. The patient thus encounters a care delivery model that is confusing to navigate, inconvenient, and severely fractured between care delivery settings. Improved coordination of care for a patient can help (see the section in this volume about “Transitions of Care”) but more broadly, the concept of patient-centered care extends beyond the approach to the individual patient, and includes as a tenet that patients have a voice in the design of the care delivery system itself. In what might be described as a traditional model for hospital administration, the patient is seen as a consumer of services, without any formal role as part of hospital operations. Consequently, despite the patient’s experiences with the institution, he or she has little or no ability to advocate for change, and the institution lacks the voice of the patient in the design of care processes. This is beginning to change as hospitals move to involve patients in operational activities, either by creating positions for patients or family members on existing hospital committees, and/or creating a separate “Patient/Family Advisory Council” (PFAC) function. Indeed the presence of a PFAC is now mandated by regulation in some states. An institution may have a single PFAC or multiple PFACs based on a desire for specific patient involvement in discreet service lines.

A role for patient/family participation has been described for a myriad of institutional processes, including strategic planning, facility redesign, research oversight, ethics, care coordination, education, finance, credentialing, leadership search, information technology, process improvement, patient safety, service excellence, and personnel practices. Table 16-1 demonstrates the range of involvement possible for patient-family advisors. Whether the plan is to involve patients on existing operational committees and/or to create one or more PFACs, a successful model for patient involvement in the design of care should begin with a vision and a plan that addresses a number of key issues:

• What are the goals for including a patient in this design process?

• What is the organizational model for patient involvement?

• Who from the organization will manage patient involvement?

• What criteria will be used for selecting patient participants?

• What are the expectations of the patient participants?

• What is the selection process?

• How long is a term of service, and are there term limits?

• What is the orientation process for patients to a hospital administrative role?

• What are the expectations for attendance?

• What criteria will be used to assess the performance of patient participants, and/or the PFAC committee itself?

An up-front strategy to define these issues is likely to be rewarded with a smooth functioning program for patient involvement in hospital operations.

FRONTIERS IN PATIENT-CENTERED CARE

How will we know if we are becoming more patient-centered? Developing measures of patient-centered care is an important step. One approach is to assess whether a hospital has the elements of patient-centered systems described in this chapter. A self-assessment tool created through a partnership of the Institute for Patient-Centered and Family-Centered Care and the American Hospital Association is available for this purpose (http://www.aha.org/aha/content/2005/pdf/assessment.pdf). But having the capability to be patient-centered is not sufficient; a system must also reliably deliver patient-centered care for each patient.

What does it mean to be patient-centered on an individual level? Certainly one can imagine situations where it is clear whether or not the care was patient-centered: for example, failure to use an interpreter when it was requested by the patient would demonstrate a failure to be patient-centered, whereas eliciting and honoring end-of-life care preferences would be a great example of patient-centered care. But in many situations, it may not be clear how to assess whether or not care has been patient-centered. Imagine a patient is admitted to the hospital for a viral illness. Despite its etiology, the patient requests an antibiotic for his illness, saying he believes it will help him improve. But instead of prescribing the antibiotic, the doctor listens to the patient’s concerns, tries to ensure he feels heard, respectfully counsels him about the risks of antibiotics, prescribes other medications for the patient’s symptoms, and reassures him that his symptoms will improve soon. In this situation, who is the judge of whether the care was “patient-centered”? Although most readers would agree that the doctor provided excellent medical care, the patient may feel that his care was not “patient-centered” because his request was not honored.

Measures of patient-centered care should be defined based on both what matters most to patients and what is consistent with acceptable medical practice. As a starting point, future work could begin to define and measure specific elements of patient-centered care, and through collaboration with patients and providers could identify elements around which there is strong consensus. Applying these measures to populations could reveal the degree to which a system delivers patient-centered care. Such information could reveal the opportunities for improvement and drive innovation by patients, providers, and institutional leaders.

Much work remains to fully define best practice in patient-centered care, and the known steps to achieving optimally patient-centered care are numerous and complex. Yet in spite of these challenges, as providers, it is our responsibility to prioritize patient-centered care. As a historical figure once said: “It is not incumbent upon you to complete the work, but neither are you at liberty to desist from it.” [Avot 2:21, attributed to Rabbi Tarfon.]

ACKNOWLEDGMENTS

Thank you to the following Beth Israel Deaconess Medical Center Patient/Family Advisors for their insightful and formative comments on this chapter: Terri Payne Butler, Peter Tarsa, and Nicola Truppin.

SUGGESTED READINGS

INTRODUCTION

Assessing and improving quality depends upon good measurement. This chapter will discuss basic principles of measuring quality in medical practice and how data can be used to identify opportunities for improvement and support improvement efforts.

ASSESSING QUALITY THROUGH MEASUREMENT

PERFORMANCE ASSESSMENT FOR EXTERNAL COMPARISON

At all levels of the health care industry, from national to local and from health systems to individual practitioners, performance assessment is now routine. Care is assessed in multiple domains, such as clinical outcomes, adherence to standards, the patient experience of care, and the cost of care. Through external comparison, or “benchmarking,” payers, the government, the public, and others attempt to judge the quality of health care providers and organizations. Performance measurement is more and more being tied to payment in the form of value-based contracts, with financial risks for performance shifting toward the providers of care, behooving us to understand how performance is assessed and improved. In order for external comparisons to be meaningful, several criteria need to be met:

• The measure should be accurately and reliably recorded across sites. As an example, length of stay for an emergency department visit will have defined starting and ending times that are routinely captured, recorded, and reported. If the standard definitions and practices are followed across all emergency departments, then length of stay comparisons should be meaningful.

• The measure should be valid and meaningfully reflect an important outcome or process. Mortality is an example of an outcome measure that has face validity, but how it is defined will affect its meaningfulness. If mortality is measured only during the inpatient stay, it will appear better for hospitals that are effective at transferring patients to other settings prior to death. Measuring mortality over a longer term, such as 30 days, 6 months, or 1 year, may be more meaningful.

Measures like mortality, patient functional status, and unintended hospital readmissions are examples of outcome measures. Outcome measures tend to be the most salient to patients, the public, and to health care providers. When using outcome measures for external comparison, however, the validity of comparison often requires adjusting for influences on the outcome that are not determined by the quality of care provided. Factors such as age, comorbid illness, and social support often impact outcomes. Some of these factors can be identified from the available data sources and included in risk adjustment algorithms. The degree to which an algorithm can truly adjust for risk depends upon the quality of the data available for the adjustment model and the power of the model itself to reflect these influences.

Process measures reflect the steps in clinical care provided to patients. Common examples of process measures used to assess hospital performance are the CMS core measures. Generally, process measures do not have to be adjusted for clinical or demographic factors, as it is assumed that these factors have little influence on our ability to carry out the processes. To be valid for assessment, process measures must also reflect meaningful clinical care. Measurement of medication reconciliation upon admission to the hospital is an example of a performance measure of questionable validity. A provider may achieve credit for performing medication reconciliation by use of an electronic tool embedded in the health record, but if an accurate list of home medications has not been obtained, the act of medication reconciliation may not be clinically meaningful.

LOCAL QUALITY ASSESSMENT AND IMPROVEMENT

External performance measures assess only a small portion of the care provided by individual health professionals and organizations. What is measured locally should reflect the priorities of the organization and the needs of patients. Accrediting bodies, such as The Joint Commission, require that hospitals routinely assess and improve care for patients. Individuals and hospitals may be less able to perform risk adjustment on internal data compared to data submitted to external organizations, yet tracking clinical outcomes against past performance can be a valid and useful way to assess quality.

SOURCES OF DATA FOR MEASUREMENT AND IMPROVEMENT

ADMINISTRATIVE OR “BILLING” DATA

Administrative data are one of the oldest sources of data on hospital performance. To bill an external payer, hospitals must accurately identify the patient, principle and secondary diagnoses, procedures, and any supply items charged. These data are generally obtained from the hospital registration system, financial system, and medical record. Names of physicians involved in care (attending physician, admitting physician, discharging physician, procedural physician, consulting physician) are generally recorded. Administrative data will include the nature of the patient’s admission (elective, urgent, or emergent), the source of admission (physician office, emergency department, transfer from other hospital or facility, etc) and the patient’s disposition at discharge (alive or dead, discharged to home or other facility, etc). Limitations of administrative data are that it is an incomplete reflection of the clinical information, is dependent upon the thoroughness and accuracy of provider documentation, and is subject to interpretation by trained chart coders and abstractors. Despite these limitations, much of the data for external performance measures originate in administrative data sets, as these data have been available for many years in standard formats across the health care industry.

CLINICAL DATA

The source of most clinical data for quality assessment and improvement is via an electronic health record (EHR). The EHR brings together patient demographics, orders, provider documentation, pharmacy data, laboratory results, radiology reports, and any clinical information documented by nurses and other health professionals. Often the EHR is interfaced with data from separate electronic systems for patient registration, pharmacy, laboratory, and radiology. Since EHRs are designed primarily to support clinical care and patient interaction, they often do not contain all of the important operational details about patient care. For instance, movement of patients within the hospital is not well reflected in most EHRs; the hospital may have a separate electronic bed management system that can be an important source of data for projects related to patient flow.

A limitation of many EHRs is that the architecture supports the care of individual patients but may not include an easy way to view aggregate data across patient populations and time. The hospital or health system may have a clinical data warehouse which holds data from the EHR and other supporting electronic systems. Data warehouses are structured to allow for more sophisticated reporting and analysis of aggregate data, and may therefore be a better source of data for local improvement than the EHR itself.

COST DATA

Costs of care can be challenging to identify, obtain, and understand. Detailed accounting of actual costs to provide individual components of a given patient’s care is not available in many hospitals. Charges may be available, but charges do not have a direct relationship to actual costs to provide care, and serve primarily for negotiating contracts and generating bills with payers. If costs are of interest in an improvement project, the hospital finance office can often provide guidance on how to generate close estimates, or how to impute costs based on utilization of resources.

CLINICAL REGISTRIES

The hospital or clinical practices may participate in collecting data for registries of specific diseases, conditions, or procedures. Professional organizations such as the American Heart Association, American College of Cardiology, and American College of Surgeons sponsor registries where clinical data from patient care at one site can be aggregated and compared to data from other sites. Often, these data are more clinically detailed and rich than data extracted from a data warehouse or an EHR since they usually reflect the skilled interpretation of trained clinical abstractors who are able to read the charts and make judgments about the care, which is more detailed than what is usually recorded in discrete data fields in the EHR.

USING DATA TO IDENTIFY OPPORTUNITIES FOR IMPROVEMENT

Benchmarking against other organizations or established targets can elucidate opportunities for performance improvement. A hospital or health system will have identified quality targets from external benchmarking or through opportunities identified internally to improve care or value to patients. Where gaps exist between current and desired performance, there are often opportunities for quality improvement (QI) projects. Aligning QI work with these identified system priorities has many advantages. Access to data, project facilitation, change management assistance, and leadership support will be easier to obtain.

Another useful approach to identifying opportunities is to develop reports from the electronic data systems or data warehouse to assess how closely the care provided matches professional standards or best practice. For example, recommended management of diabetes and hyperglycemia for most inpatients includes basal insulin with or without corrective dosing of short acting insulin. A data report showing the rate at which patients receiving short acting insulin are also given basal insulin can be useful to assess an opportunity to improve management.

Unnecessary variation in clinical practice is now a recognized problem in health care delivery. The Dartmouth Atlas project has demonstrated this phenomenon. The greater the ambiguity in care practices, the greater the likelihood exists for variation between providers. Data reports displaying variation in costs, treatment, or diagnostic approaches between providers for the same condition may highlight opportunities for improvement.

CHOOSING MEASURES FOR A QUALITY IMPROVEMENT PROJECT

USEFULNESS

Though grounded in the scientific method of hypothesis testing, quality improvement techniques differ from research methods by attempting to improve processes and outcomes in the local environment, as distinct from the intent to prove an underlying generalizable truth. The measures chosen should be useful enough for the local improvement work and should not require the degree of rigor necessary in research projects. To be useful, measures should be easily obtained and checked frequently enough to guide the improvement work. Performance should be shared with the improvement team and others in real time, whereas research methods may call for blinding the involved clinicians to reduce bias in the results.

A BALANCED SET OF MEASURES

For most quality improvement projects, there is a clear intended measurable goal. Readmissions to the hospital within 30 days of discharge, wound infection rates, and length of stay are good examples. Achieving progress on such measured goals, however, requires complex interventions occurring within the ecosystems of hospitals and other clinical environments; as such, it is frequently useful to measure other aspects of the complex systems when making changes, and create a “balanced set” of measures. A useful framework is the Clinical Value Compass (CVC), which consists of four domains: clinical outcomes, functional status or outcomes, experience or satisfaction, and costs. Not every project will have potential measures in each of these domains, yet the CVC prompts consideration of perspectives beyond the main focus, which is often on the clinical outcome of interest. Building on the example of reducing readmissions to the hospital within 30 days of discharge, one may choose to measure the combined costs of the principal and readmission stay. Patient satisfaction with the discharge process and overall care is relevant. If feasible, a measure of patient functional status at 30 days or longer may be a good complement to the clinical outcome measure of the readmission rate. If a balanced set of measures has been chosen, it is also more likely that the team will detect unintended consequences of their improvement efforts. For example, if readmissions decrease, does the principal admission length of stay increase? The outcome measure may also be remote in time and not directly connected to day-to-day work. For example, 30-day readmissions are influenced by many factors and, by definition, at least 30 days following discharge must pass to judge the effect of any intervention on the readmission rate. Adding process measures that are relevant may be very useful, such as follow-up appointments scheduled prior to discharge, transmission of discharge information to the primary physician, and confirmation of patient learning through a “teach back” method.

A GOOD MEASUREMENT PLAN

After project measures are selected, a measurement plan is necessary to make the measures available for improvement work.

MAKING MEASURES OPERATIONAL

Measures begin at the conceptual level. An example of a conceptual measure is “the rate at which patients discharged from the hospital medicine service have a follow-up appointment scheduled prior to discharge.” To make a measure operational is to define it in such a way that different people assessing the measure would come to the same answer. To express this measure as an equation, the denominator could be defined as all patients discharged from the hospital medicine service, regardless of the recommendations for postdischarge appointments or where the patients will be seen. The numerator could be defined as only those patients discharged from the service whose follow-up appointments are within 30 days of discharge and that are present on the discharge instructions with a specific date and time. With the operational definitions in hand, measures then need a specific measurement plan. If reports can be generated out of the electronic data systems, the measurement plan would include setting up those reports and validating their accuracy by comparing to the chart documentation. A simpler method in many cases is to measure a random sample. A measurement plan for sampling might include creating a work sheet for data gathering and outlining the sampling strategy, such as a convenience sample of 10 consecutive charts, or two random discharges each day of the week.

BUILD MEASUREMENT INTO THE WORKFLOW

For measures that cannot be easily pulled from electronic systems, the measurement plan must identify the collection method. Even enthusiastic improvement team members do not relish extra work. Assigning cases for retrospective chart review may meet more resistance than building measurement into the routine work of team members. Consider whether there are tools already in use that can be modified to collect the additional data. In the readmission example, does the team already work from a daily patient list on which they could record whether discharged patients have had appointments scheduled? Does an assistant already print discharge paperwork from which the follow-up appointment status can be recorded into a daily log?

TEST THE MEASUREMENT PLAN

Because measurement is key for success and should not be taken for granted, testing the measurement plan on a small scale is a useful step early in the project. The electronic report should be generated and validated early in the project (doing so allows you to trust future electronic reports). For manual data collection, test the collection tool on a few patients to verify that all abstractors interpret the operational definitions the way they were intended. The measurement plan should also ensure timeliness of the data. The pace at which the team can judge the results of improvement efforts will be limited by the frequency of measurement. When feasible, measuring on a daily or weekly basis allows the team to make progress quickly, compared to a monthly or quarterly measurement cycle.

DISPLAYING THE DATA

Optimally, data for improvement work are available frequently and are used to drive change. Creating a run chart, which is a line graph of the data over time, visually demonstrates variability, trends, and, when annotated with changes to the process, the success or failure of those changes. Regularly updating and sharing these results with team members and other stakeholders helps maintain momentum of the project.

SUGGESTED READINGS

CASE 18-1

Midway through a wound debridement, the scrub nurse noted that the sterilization indicators had not changed colors (indicating an inadequate sterilization process). A subsequent root cause analysis revealed that the sterile processing technician, at the end of his shift, forgot to push the button to start the autoclave. The next arriving technician did not notice that the sterilization indicator on the cart had not changed color, so he took the cart with the unsterile trays, and placed them on the shelf for use.

INTRODUCTION

In 1999, the Institute of Medicine (IOM) highlighted two studies from the 1980s, which suggested that between 44,000 and 98,000 patients die every year due to preventable medical errors. The subsequent IOM report, Crossing the Quality Chasm, noted, “The current systems cannot do the job. Changing systems of care will.” The report went on further to describe the six aims of safety, effectiveness, efficiency, patient-centeredness, timeliness, and equity. With these aims, the IOM has defined the ultimate vision for the US health care system.

The limitations of the current health care system were further highlighted by Elizabeth McGlynn’s study in 2003, which demonstrated that patients only received 55% of the care warranted by medical evidence. Furthermore, they found that the likelihood that an individual patient would receive all appropriate care was only 2.5%.

HUMAN FACTORS

THE INDIVIDUAL

A main contributor to the performance shortfall is the limitation of human performance. Table 18-1 illustrates expected human error rates in conditions of no undue time pressure or stress. Note that “under very high stress when dangerous activities are occurring rapidly,” the error rate can be as high as one in four (25%). Therefore, system designs that depend on perfect human performance are destined to fail at a very high rate. Furthermore, systems designed to function in conditions of high stress with frequent dangerous activities have a higher burden in order to ensure a favorable outcome.

As defined by the Federal Aviation Administration “human factors entails a multidisciplinary effort to generate and compile information about human capabilities and limitations, and apply that information to equipment, systems, facilities, procedures, jobs, environments, training, staffing, and personnel management for safe, comfortable, and effective human performance.” When considering human factors, it is helpful to consider the impact of the human and the impact of the system separately. Reliable systems must compensate for the limitations of human performance. In addition, organizational characteristics can negatively or positively impact human performance. When redesigning systems to improve performance, it is helpful to understand the factors that may negatively impact human performance, so that the design can account for the expected vulnerability.

It is helpful to understand typical human limitations, which allow for and encourage the creation of systems that may accommodate for such limitations. The first is human memory; on average, a typical human is only able to keep 7 ± 2 elements in their short-term memory. Systems that rely on human memory, therefore, are inherently imperfect. Reliable systems provide key information at the time when it is needed, rather than relying on memory. Humans also naturally cut corners or create “workarounds” when feeling rushed. Over time, repeated short cuts or workarounds result in a narrowing safety margin. This is described as “normalization of deviance.” The natural tendency to cut corners and the lack of detectable consequences falsely reassures the individual that they remain within an appropriate level of safety, or reliability, which reinforces future unsafe behaviors. Diane Vaughan first coined this term in describing the root cause of the 1986 shuttle Challenger explosion 72 seconds after lift-off. The cause was a failed “O-ring” on one of the fuel cells. The Report of the “Presidential Commission on the Space Shuttle Challenger Accident” described how normalization of deviance led to the disaster. “NASA and Thiokol accepted escalating risk apparently because they ‘got away with it last time.’” As Commissioner Feynman observed, the decision making was “a kind of Russian roulette… (The Shuttle) flies (with O-ring erosion) and nothing happens. Then it is suggested, therefore, that the risk is no longer so high for the next flights. We can lower our standards a little bit because we got away with it last time… You got away with it, but it should not be done over and over again like that.” Normalization of deviance occurs because of the natural human tendency to slip into believing that in spite of the short cuts, adequate safety or reliability margins remain. In health care, normalization of deviance is often a barrier when trying to implement and sustain basic and quality and safety interventions, such as the “Universal Protocol” or central line insertion bundles. Over time, as strict adherence to the protocols wane, and “nothing bad is happening,” then further slippage and workarounds ensue.

Stress also significantly impacts human performance by causing tunnel vision and filtering. This causes a loss of pattern recognition that humans use to rapidly discern complex situations. Fatigue negatively impacts human performance by impacting both short-term and long-term memory. The impact of fatigue is similar to having a blood alcohol level of 0.1%. Other significant factors that impact health care worker performance include multitasking, interruptions, and environmental factors (such a poor lighting, noise, distractions, etc). All of these natural human limitations must be factored into any system design (or redesign) intended to improve quality or safety.

THE ORGANIZATION

James Reason described characteristics that impact an organization’s capacity to support or impede an individual’s ability to function reliably. Human error can be addressed from an organizational perspective with a “person approach” or a “system approach.” The person approach focuses on the actions of the frontline staff who commit errors. The errors, it is believed, are due to flawed mental processes that can be voluntarily corrected, with enough motivation, attention, and vigilance. The institutional response is focused on correcting the variation in human behavior. Frequently, the responses focus on engendering fear (disciplinary measures, threat of litigation, retraining, naming, blaming, and shaming), so that the individual will focus more intently on the task at hand and not make a similar error. Often, new policies and procedures are written to ensure the correct behavior. In short, the person approach implicitly assumes that bad things happen to bad people.

In the system approach, however, human fallibility is a fundamental premise; human errors are to be expected. The errors are seen as consequences of inadequate system design. It is believed that most errors occur because system barriers and defenses that are “upstream” to the event, lead to the undesired outcome. The fundamental premise for countermeasures in the system approach focuses on changing the conditions under which humans work, rather than changing the human condition. When an error occurs, the focus is on how the defenses failed.

The person approach is somewhat appealing on at least two levels. It is emotionally satisfying to blame an individual for an adverse event. In addition, divorcing the unsafe act from the organization may be perceived as being in the best interest of the organizational leaders. But, these benefits come at a great cost. In 90% of aviation maintenance mishaps, the worker is found blameless. In order to improve it is important to perform detailed analysis of incidents, near-misses, and unsafe conditions. Within a “blame and shame” culture, the information is not voluntary reported, out of fear of retribution.

Conversely, the system approach recognizes human fallibility and system designs are successful in spite of human error. In reliable organizations, admission of errors and near-misses is encouraged and rewarded. Leadership in reliable organizations realizes that early detection of latent conditions that promote human error is essential to creating reliable systems.

As such, most organizations now create barriers and defenses to prevent the errors when designing systems of care. However, each barrier and defense is not perfect and has unique vulnerabilities. Reason describes this as a slice of Swiss cheese. However, these vulnerabilities are dynamic. Sometimes the holes are larger; sometimes they are located in a different place. For example, if one of the barriers is a second nurse checking the dose of insulin, this step may be less reliable on a specific day if the checking nurse was sleep deprived or distracted.

In well-designed systems, when one barrier fails, a second is able to catch the defect and the outcome is not compromised. In Reason’s model, the defect may pass through one slice of Swiss cheese, but is caught by the next slice. However, there are times when all the holes of multiple slices of Swiss cheese line up and the outcome is compromised … the patient is harmed or the process is not executed reliably.

Implicit in Reason’s system approach to human error is the importance of culture. Reason notes that high-reliability organizations have a reporting culture. It is essential for the staff to feel safe surfacing errors and near-misses. In high-reliability organizations, leadership accepts the accountability to create safe environments that facilitate successful outcomes. The staff member is accountable to make safe choices by following the processes that they helped to create. The staff is also accountable for surfacing existing and potential opportunities for defects. If all live up to these accountabilities, the leadership response to error is supportive of the staff, with a focus on identifying the source of the defect and developing a remedy to prevent its occurrence in the future.

RELIABILITY

As suggested in Crossing the Quality Chasm, new systems of care are required to achieve the level of reliability necessary to ensure that all patients receive the care they deserve. The new systems of care will necessarily need to account for human factors on the organizational level, as well as at the individual process level.

THE ORGANIZATION

The health care industry has a level of complexity that matches, if not exceeds other industries. Challenging the reliability of the system is that humans execute most key processes. The complex systems, the dynamic environment, and the human involved with process execution are major reasons for the reliability gap described at the beginning of this chapter. Principles gleaned from other complex organizations can help overcome the reliability gap.

Karl Weick described key organizational characteristics evident in highly reliable complex organizations. The organizations studied include, nuclear power plants, nuclear aircraft carriers, and commercial airlines. He describes those organizations as having mindfulness or “a rich awareness of discriminatory detail.” Individuals functioning in high-reliability organizations are aware of context, can discriminate details, and how the current situation differs from expectations. The five principles of mindfulness include:

• preoccupation with failure,

• reluctance to simplify interpretations,

• sensitivity to operations,

• commitment to resilience,

• deference to expertise.

Preoccupation with failure is a relentless focus on potential failure modes and how they can be prevented. An example in health care might be a detailed examination of an order set looking for ambiguities or potential error traps. A preoccupation with failure helps overcome the natural tendency to drift into unsafe behaviors that result in normalization of deviance.

Mindful organizations resist the tendency to normalize unwanted occurrences into expected events. This reluctance to simplify interpretations helps to maintain margins of safety and process reliability. Inappropriate simplification of the causes of adverse events, or a lack of focus on these events, over time, can result in major problems that might have be avoided if the initial interpretation had been more rigorous. For example, resisting attributing a medication error to a nurse failing to use the “five rights” because she did not focus (right patient, right medication, right dose, right route, right time). A mindful organization might ask if the process for executing the “five rights” is robust, and if there are adequate protections to ensure completion of the “five rights.”

Sensitivity to operations reflects a deep understanding of the processes at a frontline level. This reflects knowing what really happens in the messy world of reality, not what is policy, or what is supposed to happen. In response to low pneumococcal vaccine rates, leaders in a mindful organization do not assume that the nurses are not doing their job, but examine the process and appreciate that the cause is an overly complex screening tool.

Highly-reliable organizations realize that all systems can fail. But, these organizations have a relentless focus on not allowing that failure to compromise performance. A key element of the commitment to resilience is that the specifics are not anticipated. A preoccupation with failure might result in the development of a medical rapid response team to attend to deteriorating patients before they arrest. This team’s capabilities can be further enhanced through simulation exercises that expose the team to diverse scenarios. The simulations help reduce complexity by allowing team roles to be defined ahead of time rather than in the chaos of the acute event.

In highly reliable organizations, decision-making authority seamlessly flows to the person with the best information to make the decision. The deference to expertise might be exhibited on an aircraft carrier where a seaman can stop the activities on the flight deck because he sees a condition that might be unsafe for the landing planes. In the ICU, a nurse might “stop the line” if she does not see all elements of the central line insertion bundle, regardless of who is the central line insertion practitioner.

Other organizations may not have all the complexities of health care, but the five principles of mindfulness are directly applicable to health care. Organizations that exhibit these characteristics have a culture that promotes reliability.

THE RELIABILITY GAP

Karl Weick paints a compelling picture of what reliability looks like at the level of the organization. But, what does reliability look like at the level of the process? Reliability is intentional, and there are principles that guide that work. The lack of reliability in health care is multifactorial. Understanding these causes helps shape the interventions. While not exhaustive, the following three explanations highlight key barriers to process reliability.

1. While readily acknowledging human fallibility in others, many health care providers expect perfection of themselves. In addition, there is often a feeling that the only way to ensure reliability is to rely on no one else. From this high standard comes an over-reliance on vigilance and hard work. The reality of human factors, however, prevents the individual from performing reliably.

2. Individual providers tend to look at their personal delivery of health care one patient at a time. Ideally, the provider customizes a plan for the individual patient based on personal experience and the medical evidence. However, due to the limitations of human factors, and the paucity of high-level evidence for much of the care of the average practitioner, there is high variation in how patients are treated from one to the next. While understandable, this lack of standardization comes at a cost … complexity. It is no longer possible for an individual to deliver the full spectrum of reliable care to their patient. Teams are essential for the delivery of reliable care. The more variation from patient to patient in a care plan, the greater the level of complexity for the rest of the care team. The consequence of this unwarranted variation and increased complexity is lower reliability. This is not to say that standardization in a “cookie cutter” approach is required to reduce complexity. Rather, standardize what is “standardizable.” This relieves the care provider of the mundane and allows focus on those parts of the care plan that does require additional focus or expertise. For the rest of the care team, their work is less complex because of the reduced unwarranted variation and they are in a better position to plan and anticipate.

3. A third reason for the reliability gap is that many current processes fail to account for human factors. Process design based on human infallibility is inherently unreliable. The following section describes an approach to process design that facilitates accounting for human factors.

MEASUREMENT

The framework in Table 18-2 describes specific interventions associated with predictable levels of reliability. Interventions that result in 10^–1 reliability rely on vigilance and hard work. Examples of these interventions include:

1. Common equipment, standard order sheets, multiple choice protocols, and written policies/procedures,

2. Personal checklists,

3. Feedback of information on compliance,

4. Suggestions of working harder next time,

5. Awareness and training.

Processes that result in 10^–2 reliability use principles based on human factors and reliability science. These interventions tend to be more resource intensive. They include:

1. Decision aids and reminders built into the system,

2. Desired action the default (scientific evidence),

3. Redundant processes utilized,

4. Scheduling used in design development,

5. Habits and patterns known and taken advantage of in the design,

6. Standardization of process, based on clear specification and articulation.

THE PROCESS

The reliability design strategy was described by a group at the Institute for Healthcare Improvement (IHI). It consists of four steps:

1. Segmentation,

2. Standardization,

3. Detection and mitigation,

4. Redesign the process based on the defects identified.

1. Segmentation: When embarking on an improvement project it is often helpful to divide the population into smaller groups in order to simplify the tests of change. Process improvement is taking general knowledge or procedures and making them relevant in the local context. Unlike research, it is not possible to remove confounding variables. However, with segmentation the confounding variables can be controlled until the appropriate time.

   To illustrate the power of segmentation, consider implementation of congestive heart failure (CHF) discharge instructions. The improvement team might consider segments based on familiarity of the medical and nursing staff with the care of CHF patients. Clearly, there is a greater level of comfort on the cardiology floor than on the orthopedic floor. One might consider a continuum of comfort as follows: cardiology floor, general medicine floor, general surgery floor, and orthopedic floor. Segmenting this way, and choosing to start the work on the cardiology floor allows the team to focus on the process of delivering the discharge instructions without having to manage the barriers of knowledge deficit and comfort with the medications. Once the process for delivering the instructions is well defined and operational, the team can move on to the other floors and directly address the barriers unique to the other floors, rather than trying to address the barriers while also determining the correct process for delivering the instructions.

   The segments should have the following qualities:

   • Be based on a design theme that helps simplify the improvement activity (eg, knowledge, geography, willingness to participate, patient characteristics, etc).

   • Contain a reasonable volume, so there are enough opportunities to do tests of change.

   • Have clear-cut defined boundaries so there is no confusion about the population of patients being addressed.

2. Standardization: Standardization yields several key benefits. First, it improves reliability by reducing complexity, as described above. Second, it helps provide an infrastructure where roles and responsibilities are clearly defined. A standard infrastructure also allows for simplification of training, and competency testing. Standardizing key processes allows consistent implementation of evidence-based medicine. Lastly, it simplifies identification of defects that can be analyzed for redesign, and facilitates performing tests of change. The standardization step should be refined by serial tests of change. The reliability goal for the standardization step is 80%. If the step is less than 80% reliable, the defects will overwhelm the detection and mitigation step, which tends to be more resource intensive.

3. Detection and mitigation: The value of the detection and mitigation step is that it helps reduce the complexity of the standardization step. Since the standardization step need only capture 80% of the opportunities, it is not necessary to develop contingencies for lower frequency occurrences. If the standardization step had to include those contingencies, it would be too complex and therefore less reliable. The detection and mitigation step includes two processes. The first process reliably identifies the defects from the standardization step utilizing 10^–2 level interventions. The second process mitigates the defect, using 10^–1 level interventions. An example might be detecting failures by running an electronic report daily for all patients who have not been screened for administration of the pneumococcal vaccine. Those failures would then be mitigated by a supervisor contacting the nurses caring for the patients who experienced the process failure.

4. Process redesign based on defects identified: In pursuit of continuous improvement, the fourth step in the reliability design strategy is to examine the defects identified in the previous step and feed the learning back into the design of the standardization and detection and mitigation steps. Once the processes in the first segment achieve an acceptable level of reliability, the team should move on to the next easiest segment. Now that the processes have been defined and tested, work in the next segment can focus on the barriers unique to the new segment. Attacking the segments from easiest to hardest allows the team to gain experience and further refine the process, which simplifies the work in even the most difficult segment.

CONCLUSION

As highlighted by the IOM reports and McGlynn’s work, patients are regularly experiencing preventable harm and are not receiving all of the care that is intended. Closing that reliability gap requires focused multidisciplinary team involvement to develop and oversee the process redesign. Understanding human factors and having a structured model for increased reliability is essential. Reliable process design, regardless of the improvement methodology, is an effective approach to closing the reliability gap.

SUGGESTED READINGS

ONLINE RESOURCES

INTRODUCTION

Every organization needs a structure and a tool kit to support employee and patient safety and continuous quality improvement. Since what cannot be seen cannot be fixed, robust identification of adverse events and latent safety threats (risk to patients, families, and employees) should be a priority of every hospital. Here, we discuss structures and tools to identify adverse events and risk. Once identified, the hospital and staff must then prioritize efforts and determine which techniques will be applied to reduce adverse events and risk. Risk reduction efforts may focus on error prevention or on harm prevention, and both approaches are key to risk reduction. Table 19-1 reviews definitions for the key terms used throughout the chapter.

THE ROLE OF THE CULTURE OF SAFETY IN IDENTIFYING RISK

In organizations with a robust culture of safety, it is evident to all involved that safety is the top priority. People working in the area focus on safe practices and support one another in delivering safe care. Individuals in a mature culture of safety exhibit the principles of a highly reliable organization, including “preoccupation with failure” as described by Weick and Sutcliffe, such that there is a constant obsession with mitigation of risk. Instead of ignoring small, nagging concerns, workers share those concerns with others, and team members rally to help resolve the concerns. When the culture of safety is strong, people report concerns without fears of retaliation. Employees respond supportively to concerns about risks to patient safety and do not seek to blame individuals when an error occurs. When safety is a priority, leaders routinely inquire about safety concerns and take the time to listen, seek to understand root causes, demonstrate their commitment to safety through action, and communicate back to staff on the organizational response to adverse events and concerns. An organization’s culture of safety can be measured by the Agency for Healthcare Research’s survey on patient safety culture. The scores of this survey have been correlated with patient outcomes in multiple studies and can assist an organization with monitoring their culture over time. Specific questions may also be useful to assess the effect of culture on identifying problems (Table 19-2).

RISK IDENTIFICATION AS PART OF A SAFETY PROGRAM

Identification of patient safety risks requires an intentional multifaceted approach. The National Quality Forum identifies and promotes safe practices in health care (see http://qualityforum.org). Among these best practices is the presence of leadership structures and systems to ensure organizational awareness of safety failures and performance gaps that need attention. Government and commercial health care purchasers, insurance providers, and hospital accreditation organizations provide strong incentives for hospitals to invest in infrastructure supporting the identification, analysis, and correction of risk. Less obvious sources of harm, including deviations from generally accepted practice standards such as evidence-based guidelines of care, are also important to detect. For example, identifying variation within a subspecialty of blood product ordering, compliance with tonsillectomy guidelines, or ordering of computed tomography scans for suspected appendicitis is likely to lead to institutional cost savings and improved quality of care. Because information about errors and risks comes from many potential sources, hospitals face challenges to bringing this disparate information together, investigating and mitigating risk, and providing feedback to employees about actions taken in a meaningful way.

One major challenge is that only a small fraction of preventable harm is detected by most methods, especially those based on voluntary human reporting. Computerized data mining methods tend to greatly increase the number of potential errors identified, but the specificity can be low. For example, a rule to identify nephrotoxicity from medications may look for a rise in the serum creatinine during hospitalization. Most of the cases identified will not result from an adverse drug event, and additional resources will be required to further investigate each case.

Knowing that the detected errors are but a sample of true preventable adverse events, hospital safety leaders must decide whether there is value to applying additional resources to enhance detection. Bringing together the multiple sources of information allows for a broader view, and this effort benefits from having a central safety committee or safety officer to whom the information funnels.

METHODS OF IDENTIFYING ADVERSE EVENTS AND ERRORS

Some methods are reactive and retrospective, being generated in response to specific events that come to attention. In contrast, systematic methods tend to identify latent or hidden errors or risk points, and are often both reactive and proactive. Both types of approaches are necessary in the overall strategy to identify risks. Certainly, there are some errors that will never come to the attention of leadership, either because they are not recognized as errors or because employees do not feel comfortable reporting them.

REACTIVE METHODS

Event-reporting systems

Voluntary event-reporting systems rely on employees to take time to report a concern, trusting that action will be taken by leaders, and that retaliation will not occur. In a healthy culture of safety, employees report freely and openly, with few barriers. Employees do not fail to report events that seem to represent temporary or minor areas of potential harm, because they are aware that these are important opportunities to learn. Indeed, a robust reporting system will collect a significant number of system failures or errors that were successfully intercepted or that did not result in harm, known as near misses. Near misses are golden opportunities to identify risk-prone conditions or processes and to intervene before harm results. The ideal state would involve occurrence of near misses only, and no incidents of preventable harm. An effective reporting system enhances the engagement of frontline staff in patient safety by providing an identified channel for their observations. To be effective, employees must be aware that the reporting system exists, reporting must be user-friendly, and it must be known that leaders place value on the reports. Timely acknowledgment, expressions of appreciation, and feedback regarding actions taken reinforce the desired reporting behaviors.

Ease of reporting is key to maintaining a low reporting threshold. Paper reports and verbal reports via telephone recording have advantages of speed, though the information has to be transcribed and aggregated separately. Lengthy written or electronic reporting systems with a large number of required fields are not likely to be used widely. Electronic reporting systems may prompt for more precise and complete information from each report and may produce structured reports from which data are more easily analyzed. Electronic systems may also enable immediate notification of appropriate personnel. For instance, an event reported as causing significant patient harm may generate an automated communication to a risk manager, safety officer, or hospital leader, facilitating a timely response to the event. Of note, most states provide legal protection for event reporting such that the information contained with the reports is not discoverable.

How leaders and managers respond to aggregate data from event-reporting systems will send strong signals to employees. Because event reports are dependent on willingness to report and are unlikely to reflect true incident rates, leaders and managers should exercise caution in inferring that a high number of events reported represents worse safety in one area versus another. In fact, the number of reports may be more indicative of the culture of safety than of safety itself. However, it is human nature to conclude that higher numbers of reports indicate an area of unsafe conditions. Managers may worry that reports reflect poorly on their performance and discourage use of the reporting system. Leadership, particularly of the “C-suite” and the board of directors, should pay attention to the use of data for learning rather than for judging in order for the reporting system to be effective.

The options for reporter identity protection deserve intentional thought when designing an event-reporting system. An open system makes no attempt to protect the reporter’s identity, so colleagues and supervisors may identify the reporter. This kind of system may work in organizations where the culture of safety is strong enough that there is no retribution for reporting and, in fact, reporting is rewarded, whether by peer appreciation or formal recognition. It is worth noting most health care organizations are not yet at this point in their safety culture journey. If reporters face criticism or retaliation in even a few instances, willingness to report may be severely affected.

A confidential reporting system allows identification of the reporter only to responsible system administrators who will follow up on the event with the reporter. Confidential reporting may overcome reluctance of some people to report and enhance detection of sensitive issues. For example, a person reporting inappropriate sexual comments or behavior may be reluctant to have his or her identity known to the person whose behavior is being reported.

Anonymous reporting serves to fully protect the reporter’s identity and thus may expand reporting of sensitive events and reporting in work areas where a climate of fear exists. If inadequate information was submitted to identify the event, however, further investigation and learning are severely hampered. To the other extreme, if the reporter gave full information in a detailed report and few people are fully aware of the event, it may be impossible to maintain anonymity. Regardless of whether reporting system is open or confidential, it is important in most health care organizations to allow for the option to report anonymously. Where necessary to provide for anonymous reporting, efforts to improve the culture of safety will ideally allow movement toward a more open reporting system.

PATIENT COMPLAINTS AND CONCERNS

Patient complaints and concerns are another source of reports, whether submitted by patients or families or by employees on behalf of them. The perspective of patients and families may be very valuable and complementary to the insights of staff members. These reports are more likely to reflect the level of service, compassion, communication, and partnering with patients and families. The reports are not likely to be anonymous and bring with them a duty to respond back to the patient or family. As with reports from staff members, treating the information as a learning opportunity may shape how frontline workers respond to complaints. While the volume and content of patient complaints and concerns may predict the likelihood of legal action, it may be difficult to draw actual conclusions about patient safety from rates of complaints.

CLAIMS ANALYSIS

Analysis of medicolegal claims may be a tempting source of information, but it is less likely to generate useful ideas about how to improve patient safety, compared to other methods. Since the vast majority of patients harmed through medical care do not bring claims, this subset is idiosyncratic. Deep understanding of any given event degrades rapidly with time, and investigation of an event should have occurred long before a claim is filed.

SYSTEMATIC METHODS

Patient safety walk rounds

Scheduled rounding in patient care areas by leaders may be an effective method to accomplish several goals, including promotion of a culture of safety. As described by Frankel et al, structuring the content of rounds and recording the comments and concerns of staff members may yield valuable insights. Past events deserving of investigation and concerns about ongoing risks can be heard directly by the leadership team. These reports stimulated by the visit of leaders may not have been collected through other means. It is important to provide feedback to employees on items discussed in safety rounds. If possible, it is useful to develop a rotating schedule to include all shifts and all areas.

MORBIDITY AND MORTALITY CONFERENCES

Morbidity and mortality (M&M) conferences are a time-honored tradition in medicine. Discussion of patient deaths, complications, or harms resulting from care provides a learning opportunity for the attendees. Too often, perhaps due to the confidential nature of many conferences, the learning stops at the door, and feedback to the larger system does not happen. Traditional M&M often involves individual blame rather than a systems-based approach. Additionally, the discussion of failures may be inhibited if the culture is not open, and the lack of a structured process of case review may lead to wide variation in the conclusions drawn from case review.

Structuring case reviews, identifying underlying causes of failures, involving organizational quality and risk leaders, and assigning responsibility for making system changes can make the M&M conference benefit the larger system.

TRIGGER TOOLS

Adverse events and errors that surface through reporting systems and complaints poorly represent the rate at which such problems actually occur. Structured reviews of medical records can identify problems that have not been reported and provide a rate estimate. One such method has been developed by the Institute for Health Care Improvement. Triggers are explicit criteria or clues to the presence of an adverse event, stimulating a deeper review of the record. For example, administration of naloxone as a trigger is likely to detect some causes of nonpreventable harm as well as preventable harm. Errors in warfarin management may be identified among patients in whom vitamin K is administered. Most commonly trigger tools are used to detect adverse drug events, but they can also be used to detect other causes of potential preventable harm. Applying the tool to a random sample of hospital discharges can give an estimate of the rates of common sources of harm to patients. These rates can be tracked over time, and the information obtained may be used by the hospital to set improvement priorities.

MONITORING OF HIGH-RISK PROCESSES

Some processes of care are inherently risky, such as provision of sedation, performing invasive procedures, and responding to cardiopulmonary arrests. Particularly concerning are the low-frequency, high-risk events. Monitoring the processes of care and the outcomes may identify deviations from standard practice that may pose undue risks and point to opportunities for improvement. Using the example of cardiopulmonary resuscitation (CPR), the hospital may have a mechanism to review that the right complement of personnel responded to the emergency, that the team was alerted in a timely fashion, that a team leader was identified and roles assigned, that the patient was correctly assessed, that the correct resuscitation algorithms were followed, and that the medications and other therapies were correctly administered. The results of CPR reviews and the outcomes of the resuscitation attempts are then reported to the appropriate hospital representative or committee. Debriefing immediately after such events is likely to yield discovery of areas for improvement.

In a hospital with a robust culture of safety, the insights from monitoring are shared and lead to actions to address deficiencies. Suppose, for example, resuscitation monitoring identifies that patients with difficult airways are not managed as well late at night as during the daytime hours when an anesthesiologist is available. This insight should be reviewed by hospital leaders who determine the actions needed to provide better airway management at night. One solution may be to train and certify hospitalist physicians to manage difficult airways. Additionally, processes may be monitored using open auditing such as observing the surgical safety checklist preoperatively or using secret shoppers posing as family members to audit hand-hygiene practices. Health care processes must be monitored using multiple different methods to assure compliance.

IN SITU SIMULATION

Both Institute of Medicine reports, To err is human and Crossing the Quality Chasm, suggested that health care utilizes both crew resource management and simulation to improve patient safety. In situ simulation involves practicing team-based care on actual patient care units and this form of training provides a method to identify latent safety threats.

MINING ELECTRONIC DATA

Health care facilities are rich in electronic data that can be mined for possible adverse events. Every hospital has administrative systems to support billing. Among the diagnosis codes are those that may indicate safety problems, such as codes for accidental puncture or laceration of an organ and for foreign body left in during a procedure. Since October 2008, hospitals routinely indicate whether conditions were present on admission. Reviewing diagnoses such as pressure ulcers and deep vein thromboses that were not present on admission may identify opportunities for improvement.

Electronic laboratory and pharmacy systems can be used to identify potential adverse events for further review. Acute renal failure occurring after admission to the hospital may be identified by searching the laboratory data for patients in whom the serum creatinine has risen significantly. Practically, however, often the needed data are available within the electronic health record, but there are not enough resources available to satisfy the many demands for data throughout the system.

FAILURE MODES AND EFFECTS ANALYSIS

Failure modes and effects analysis (FMEA) is a prospective technique to anticipate the ways in which a process may fail and to prioritize the efforts to prevent failures. With roots in the military, FMEA is widely used in manufacturing and more recently is being applied to health care delivery. While there are a number of models, including health care failure modes and effects analysis used in the Veterans Health Administration, FMEA generally follows these steps.

1. Choose the target for analysis. Because conducting FMEA requires expertise and the commitment of significant time and resources, a hospital must select where to apply the technique. A frequent consideration is to focus where prior problems have occurred. For example, an institution might use FMEA to identify and prevent errors in the placement and use of nasogastric tubes. FMEA can help a hospital safely implement a new technology such as bar coding the steps of medication preparation, distribution, and administration.

2. Assemble the team. To understand the process in fine detail, it is critical to identify and involve team members with expertise from all disciplines that use, interact with, or maintain the process. A trained facilitator should be part of the team, as FMEA has a specific structure.

3. Describe the process to be analyzed in detail. This is best done graphically, beginning with a high-level flow diagram that serves as a framework for analysis. The flow diagram allows team members to share a mental model of the process steps. From the high-level diagram, the team can develop a detailed understanding of the different steps and supporting processes. When important gaps in the process cannot be filled in by team members, additional information is sought.

4. Identify the ways that failure may occur (failure modes). Drawing upon the expertise of team members and upon data where available, the team identifies vulnerabilities and underlying causes among the process steps. Failure modes may occur under normal conditions or when the system is stressed.

5. Prioritize the failure modes. The team must decide where to focus attention for improvement or process redesign among the failure modes found in the previous step. To prioritize, most FMEA models apply a grading scale (eg, 1-10) to several aspects of each failure mode and its causes:

   • Severity: The effect of each failure is described and assigned a grade. A low grade indicates a failure that would have minor consequences or is easily recoverable; a high-grade failure effect is catastrophic (would cause grave harm and cannot be stopped once the failure occurs).

   • Frequency: The probability of occurrence of the failure mode or its causes is also graded. Low-grade frequency suggests a very rare event and high grades suggest more frequent events.

   • Risk of escaping detection: Failures that are immediately obvious allow for early detection and the opportunity to recover or to mitigate the effects. Easily detected failures receive a low score for risk of escaping detection. Other failures may be detected through routine inspection some time after occurrence. The highest score goes to failures that are not detected until the outcome (failure effect) has happened.

   • Multiplying the grades of severity, frequency, and risk of escaping detection for each failure mode produces a risk priority number. The failure modes can then be ranked, with the highest risk priority numbers indicating where the team should focus attention. Because this ranking process is imprecise, the results provide a guide rather than a prescription for the next steps.

6. Determine the action steps. If the FMEA leads the team to the conclusion that the process is far too unsafe to continue, a complete redesign may be necessary. Or, the team may conclude that bringing in a new piece of equipment is unjustified given the associated risks. More commonly, the team will identify ways to eliminate causes of failures, to provide earlier warning of failures, or to mitigate the effects of failures in service of improving overall safety.

RETROSPECTIVE INVESTIGATION OF EVENTS ROOT CAUSE ANALYSIS

Root cause analysis (RCA) is complementary to FMEA. It is a retrospective technique that provides a robust structure to review an adverse event or near miss. RCA is reactive; FMEA is proactive. But, the ultimate goal of each technique is to identify ways to prevent future adverse events. See Table 19-3 for a comparison of these two techniques.

Effective RCA requires a detailed, intimate understanding of the event being studied and necessitates the participation or interview of people directly involved in the event. Because RCA is usually performed in response to a recent adverse event, emotions of people involved may be fragile. A poorly performed RCA that permits blaming of individuals can seriously undermine the culture of safety. RCA done well results in learning the underlying (root) causes of human failures, process failures, or equipment failures. Focus on correcting the root causes of failure fosters safer care of patients and a safer work environment for employees.

Conducting RCA requires a trained facilitator and an investment of time and resources. Hospitals will have to select where to focus this tool. Serious adverse events, such as the unexpected death of a patient related to an error, are obvious targets for RCA. However, near misses that reveal a potentially serious process failure should also be considered for RCA. A patient who recognizes that his chemotherapy has been mixed incorrectly because it is the wrong color may have prevented a serious medication error. RCA of this error may reveal weaknesses in the chemotherapy mixing process that need to be fixed. The Just Culture algorithm would encourage us to handle investigation of events independent of the outcome that resulted. Failure in duty to follow a procedural rule would result in the same treatment of the employee whether the patient died or caused no harm. Root cause analysis will generally follow these steps:

1. Assemble the team. Some hospitals have an existing team for conducting RCAs. This team rarely has firsthand knowledge of the event being reviewed and will probably not have an intimate understanding of the focused work processes and environment of those involved in the event. The team will add members with such knowledge or will gain that knowledge through extensive interviews. More commonly, a team is brought together specifically for the RCA. In addition to the trained facilitator, the team should be interdisciplinary and involve people who work closely with the processes being evaluated. Some experts advise against including people directly involved with the event as team members because of their potential difficulty with being objective and open in such a forum. If those with direct involvement are invited to join the RCA team, the facilitator must be sensitive to this conflict and create an atmosphere conducive to openness. If employees who were directly involved in the event do not participate in the RCA, they should be interviewed in depth in order to understand the processes, culture, and root causes. Hospital leaders, when possible, can strengthen team function by participating and supporting improvement opportunities that result from the RCA. If a leader does not participate directly, knowing that a leader will closely focus on the team recommendations can also lend weight to the RCA.

2. Set the atmosphere. In teams brought together for the RCA, the leader or facilitator may be the only person formally trained or experienced in the technique. Setting the atmosphere by providing an orientation to the process and laying out ground rules for the conduct of the RCA can be critical to success. Among the ground rules should be a prohibition against finger-pointing and personal attacks. While humans may have erred, directing blame at an individual stifles learning. Rather, for every failure point, ask “why” and “what conditions existed to permit this failure?” Human error alone should not be an acceptable root cause. Another useful ground rule is to avoid speculation. Where gaps in understanding occur, the team should seek additional information. It is also important to discuss up front whether peer review statues will protect the information discussed, and whether or not the information will be discoverable.

3. Describe the events. Create a detailed flow diagram or a timeline of the sequence of events. The team seeks additional information to fill in gaps in understanding.

4. Identify root causes. Using structured questions as in the triggering and triage questions of the VA National Center for Patient Safety (see http://www4.va.gov/NCPS/rca.html) creates a more complete analysis by prompting consideration of categories of causes, including environmental conditions, equipment function, policies and procedures, training, communication, and fatigue. Again, digging deeper with each question and not accepting human error or procedural violation as a root cause are essential to identifying preventable causes.

5. Identify solutions to prevent recurrences. Using standardization and reliability science will create more robust actions. Consider where similar vulnerabilities exist in the organization and generalize the learning where possible.

6. Report findings and recommendations to leaders. This step will help secure leadership support for actions needed.

7. Follow-up. It is important to complete the action items discussed in the RCA within a specified time period. Teams may consider stressing the system using in situ simulation to determine if the same vulnerabilities still exist or if the solutions implemented have closed the gaps.

INCIDENT INVESTIGATIONS

Most adverse events and near misses will not be investigated with an RCA, simply because the hospital is not capable of responding to all with this level of investigation. Still, the personnel performing the investigations should be seeking root causes of problems, detecting trends, identifying solutions, and generalizing to other areas. The results of incident investigations should feed back into the safety structure of the hospital through the safety committee or safety officer so that further action may be taken. Providing feedback to reporters is a key.

PEER REVIEW

Peer review processes are intended to adjudicate the competence and professionalism of health care providers. As such, peer review is fundamentally different from the investigations already described. This approach focuses not on problems in the health care system, but on individual performance. Peer review usually occurs as a physician or other provider is newly hired or granted specific hospital privileges to confirm competence or in response to some concern raised about performance. Concerns may arise in response to a particular incident or in response to data compiled over time. For instance, an unexpectedly high rate of procedural complications or resource utilization compared with peers may prompt closer review. Using a construct such as Just Culture may facilitate a fair treatment when the temptation is to punish the employee in the face of a bad outcome.

Peer review may generate insights about an error-prone process or other system problems that can be fed into the hospital safety or quality improvement structure. Judgments about the professional competence of individuals are managed through hospital medical staff governance structures.

As a strategy for reducing errors, peer review follows a “person approach” that is weak to the extent it assumes that human competence and behavior are responsible for most errors. Reliance on human perfection flies in the face of what is known about human performance in a variety of tasks, and the personal approach to error reduction undermines a culture of safety when individuals are blamed for errors and outcomes. The goal of peer review should be to ensure that health professionals are able to function on a level commensurate with the specific privileges granted. A “system approach” to error reduction assumes that most errors result from flaws in the system in which people work. The focus turns to creating a system that recognizes human fallibility, discourages and does not facilitate workarounds, and prevents harm to patients.

Peer review also encompasses the behavioral norms expected of health professionals. Behavior that is disruptive to safe patient care or that threatens the safety of employees, such as verbal abuse of employees, lying about events in the care of patients, throwing public temper tantrums directed at others, or making unwanted sexual advances, must be dealt with by leaders. Failing to do so contributes to fear among employees and distrust of leaders, undermining the culture of safety.

RESPONDING TO IDENTIFIED ERRORS AND RISK POINTS

PRIORITIZATION

Health care delivery is an inherently complex and dangerous field. Patients enter hospitals with conditions that may be either known or unclear. The methods we apply to diagnosis and treat patients involve sharp objects, ionizing radiation, and toxins, delivered in a team setting where we must plan and communicate clearly across professional disciplines and across multiple handoffs. The challenge then lies in deciding which problems to tackle first and which problems to set aside. When creating new processes or introducing new equipment, leaders must take the opportunity to apply FMEA or other strategic safeguards so the number of potentially chaotic and unsafe processes does not continue to rise exponentially.

Frontline employees bring their expertise about the risks they encounter. They should collaborate with the safety officer or committee that has a broader view of the problem areas in the hospital. The broader view comes from formal assessments of high-risk areas or processes, as with the FMEA technique, and from the accumulated experiences of risks and errors collected from the reporting system and other methods already discussed. When appropriate, local problems can be addressed with local solutions by the people who do the work. The hospital safety officer or committee will decide where and how to apply the more resource-intensive efforts to improve safety.

AVOIDING OVERREACTION AND UNINTENDED CONSEQUENCES

In the face of an adverse event, it can be tempting to apply intense focus and impose a quick solution. Indeed, when a serious continuing risk of harm is found, an immediate safeguard is appropriate. Reacting quickly to every danger, though, may cause loss of focus on higher-priority risks. Ill-considered solutions can overburden employees to the point of paralysis or can introduce additional harms. As an example, computerized order entry systems permit checking for drug-drug interactions as a safety feature. System administrators can select the level of interaction (severe, moderate, minor) at which an alert interrupts the prescribing process. If, in response to an adverse drug event, the level is set to include all potential interactions, the number of interruptions may overwhelm prescribers, creating alert fatigue. As a result, cognitive errors may increase and prescribers may ignore more serious drug alerts. The net effect may well be to increase risks.

CONCLUSION

A robust safety program helps to manage the complex and dangerous inpatient hospital environment. It sets strategies and tactics for engendering a safe culture, identifying and investigating harms and risks, and prioritizing improvement efforts. The tools described in this chapter support the execution of such a robust and comprehensive safety program.

ACKNOWLEDGMENT

The author would like to acknowledge Nathan Spell, MD, for his contribution to the first edition chapter.

SUGGESTED READINGS

INTRODUCTION

In November 1999, the Institute of Medicine (IOM) issued the report To Err is Human, detailing a problem of preventable medical errors that were killing as many as 98,000 inpatients per year. Subsequent publications have estimated it may be as high as 400,000 per year. Specific types of medical errors highlighted in the IOM report include error in the administration of treatment, failure to order and follow-up on indicated diagnostic exams, and avoidable delays in care and treatment. Many years later problems still exist: nearly 2 million patients a year develop infections during their hospitalizations, and 90,000 to 100,000 of those infected die, while hand-hygiene rates range from 30% to 70% at most acute care facilities. The IOM report also estimated that medical errors cost the US $17 billion to $29 billion a year, and called for sweeping changes to the health care system to improve patient safety.

Improvements in patient safety have focused on addressing the root causes of these preventable patient harm events, specifically events related to poor communication, lack of teamwork, fragmentation of care, and a lack of leadership from the medical community. In addition, patient safety experts have also implored physicians and hospitals to approach patient harm events with transparent, open, and honest communication between caregivers and patients and families in order to learn from mistakes and poorly designed systems.

This chapter reviews ways in which hospitalists may actively participate in the prevention of patient harm and provide appropriate management and assistance when patient harm does occur.

PREVENTING ADVERSE PATIENT EVENTS

Most patient safety experts would agree that the areas of highest priority to proactively maximize patient safety fit into three broad domains: communication, teamwork, and leadership. Within each of these domains lie critical concepts and issues about which the highly reliable and safe-practicing physician must remain mindful.

COMMUNICATION

No chapter on the prevention of patient harm is complete without a major focus on the role communication—or lack thereof—plays in serious patient safety events. The most common types of communication of high priority in patient safety are listed in Table 20-1.

Handoffs

Year after year The Joint Commission (TJC) publishes data showing 65% to 70% of all sentinel events are rooted in communication breakdowns. It appears that since the implementation of the Accreditation Council for Graduate Medical Education resident physician work hour limitations the communication problems have increased, especially in the area of handoffs, when the responsibility of care is passed from one provider to the next. With this limitation of resident physician work hours, the need and demand for hospitalists to “fill the gaps” in patient care has increased substantially. Associated with that increase in demand, hospitalists in particular have recognized the imperative of a standardized, user-friendly, and reliable method of handing off care from one provider to the next.

The content and process for handing off in the inpatient setting has evolved as practitioners try to meet regulatory requirements while maintaining simplicity, efficiency, and usability of the various handoff tools that are available. Various pneumonic tools, such as Situation, Background, Assessment, and Recommendation have been devised to assist in the handoff process but have come and gone from institutional policies and guidelines as providers struggle with a reliable way to meet this important imperative. Hospitalists must play a role in designing and implementing a best practice handoff process appropriate for the context in which they work.

Vendors of electronic health records (EHRs) have also entered the arena with EHR-based tools to facilitate the often onerous process of handing off care of large numbers of patients. Regardless of the chosen method, all hospitalists must employ a reliable process to transmit necessary patient information from physician to physician. Defective or unreliable handoffs substantially increase risk of patient harm and the associated liability.

Documentation in the electronic health record (EHR)

With the passage of the 2010 Health Care Reform Act, it has become increasingly clear that the use of electronic health records will become much more ubiquitous in the coming years. While patient safety benefits of EHRs are well documented, only recently have informatics experts been publishing the unintended, unsafe consequences related to their use.

One of the most glaring examples of an unintended, unsafe consequence to EHR implementation is the abuse of “cut and paste” or “copy and paste” functionality, the process by which entire sections of nursing or physician documentation are copied and pasted from past to present notes. Numerous published reports demonstrate cases in which erroneous information has propagated, almost “virally,” throughout a patient’s EHR through the use of copy and paste. This process creates unsafe conditions for the patient such as in the example of the erroneous propagation of a “faux” allergy to an important medication. Serious medical-legal consequences can result for those who continue to misrepresent medical information through subsequent “copies” of the erroneous information or for those who act upon this unreliable information. The credibility of physicians comes into question when they are forced to defend misinformation they have propagated throughout the medical record, such as a temperature of 1101.5°F or a blood pressure of 1180/60.

While the “copy and paste” functionality provides useful efficiencies for documentation of long lists of medications or past surgical procedures, hospitalists must be aware of the deleterious consequences of the inappropriate use of this functionality and they should serve as positive role models and mentors throughout the organization for others who document in the EHR. From a patient safety and legal perspective, it is also incumbent upon the hospitalist to facilitate the correction of erroneous information encountered in the EHR.

MANAGEMENT OF CRITICAL TEST RESULTS

Critical test results management cuts to the heart of the health care business. US hospitals complete approximately 12 billion diagnostic tests every year. Most test results are within normal range and do not require follow-up by the clinician. However, a small but important number of test results, approximately1% to 5% of a hospital’s test volume, are abnormal or critical. Hospitals and hospitalists have a professional, legal, and ethical obligation to ensure that these results are communicated to the responsible physician and appropriate action is taken.

Traditional systems to communicate and manage critical results are full of potential points of failure. In many hospitals, especially for hospitalists, contact information changes on a regular basis. Radiology departments and the pathology lab may not have the correct contact information for the responsible physician. Faxes can be equally problematic as the receiving machine might be off or out of paper. And once communicated, the right person might not receive the information. Unfortunately, radiologists and laboratory technicians may spend hours or days trying to track down the appropriate physician for results communication. Not surprising, miscommunication of critical findings have been identified as the causative factor in 85% of radiology lawsuits. Appropriately, The Joint Commission has deemed the management of critical test results as a national patient safety priority and requires hospitals and health care professionals to improve processes involved in such results. To improve the safety and quality of care their patients receive, hospitalists must play an integral role in the design and implementation of systems and processes to manage critical test results. At a minimum, in the hospital setting they must actively participate in a process to ensure the proper identification of responsible physicians and an efficient means for involving those responsible physicians in the communication and action based upon these results.

TEAMWORK

MULTIDISCIPLINARY ROUNDS

Data abounds on the value teamwork brings to the safe and effective delivery of health care. The days of a single physician effectively micromanaging a patient’s entire hospital stay are long gone. Research has shown that physicians can mitigate the negative effects of the necessary fragmentation of health care delivery by participating in multidisciplinary rounds during which physicians, nurses, pharmacists, and other allied health professionals discuss the daily plan for the patient and coordinate the transition of care to the outpatient setting. Inclusion of the patient and family in these rounds also provides benefit.

Especially for complicated patients, multidisciplinary rounds have been demonstrated to reduce length of stay, decrease the incidence of medication errors, prevent hospital readmissions, and improve overall patient and family satisfaction related to the hospital stay. As health care reimbursement models transition to a “pay for performance” or “pay for quality” metric, hospitals and hospitalists will find multidisciplinary rounds fundamental to the business model of health care.

INFECTION PREVENTION

All patient safety and quality organizations as well as health care regulators have identified health-care-associated infections (HAIs) as a top priority. The human and financial toll of HAIs accounts for a large portion of preventable harm in the United States. Therefore, the elimination of any significant proportion of HAIs is paramount to control health care costs and preventing patient harm. The role of the hospitalist as an essential, active member of the health care team is central to the efforts to reduce HAIs.

Three easily identifiable areas of quality improvement related to prevention of HAIs include (1) hand hygiene, (2) prevention of catheter-associated urinary tract infections (CAUTIs), and (3) prevention of intravenous line-associated blood stream infections, especially those related to central lines (CLABSIs). With observed hand-hygiene rates in most hospitals hovering around an abysmal 40% to 50%, administrators struggle to find solutions. Many leaders have found the solution to the hand-hygiene dilemma in the active engagement of hospitalists in their institution-wide efforts. Numerous success stories show that hospitalist engagement, by actively promoting hand-hygiene within their team, has taken 40% to 50% compliance rates to a sustainable 90% or higher. From the patient perspective, these increases in hand-hygiene rates translate into substantial reductions in methicillin-resistant staph infections and other HAIs.

The involvement of hospitalists on daily rounds in which they are able to order the removal of nonessential Foley catheters substantially reduces the incidence of CAUTIs. The same holds true for CLABSIs, in which the reduction in the days of use for invasive intravenous lines is also associated with a heath care-associated infection. With all these efforts, hospitalists are ideally situated to affect safety outcomes for their patients and others on their teams and in the institutions where they practice.

PATIENT TRIAGE AND RAPID RESPONSE TEAMS (RRT)

As the frontline physicians accepting or coordinating inpatient hospital admissions, the hospitalist has an affirmative obligation to make certain newly admitted patients are placed on units and into beds that are appropriate for the level of care they need. Nonetheless, patients on appropriate wards or units will still deteriorate faster than the care professionals anticipate. When that happens, in the interest of patient safety, hospitals must have a process for rapidly summoning a team of professionals to assess the patient’s current state of deterioration and to assist in a “retriaging” process.

Whether activated by other physicians, nurses, patients, or family members, hospitalists play an important role in the response to the deteriorating patient in many institutions. Effective and valuable physician members of a rapid response team, or any team assigned to respond to the clinically deteriorating patient, possess certain necessary attitudes and skills. As a leader of the team, the hospitalist must approach each “call for help” with a high degree of “mindfulness” in order to avoid premature closure based upon selective information provided to them by the care professionals previously caring for the patient. Regardless of whether the “call for help” or activation of the RRT seems appropriate after the initial response and investigation, it remains critical that the physician leader of the response team supports those who trigger the activation and helps prevent ridicule or criticism of those who asked for help.

Hospitalists must remain open minded while gathering all potentially important information that might be useful for arriving at solutions or treatments to reverse the deteriorating trend in condition. While carefully listening to other team members and caregivers, the leader of the team must be able to synthesize an approach that considers all the relevant factors, especially unexpected findings, and avoid the temptation to disregard information that might be inconsistent with their preliminary diagnosis—the concept of premature closure. They must remain cognizant of their own confirmation biases and strive to keep them in check. Not all chest pain is a myocardial infarction and not all wheezing is asthma!

From the skills perspective, a successful hospitalist RRT team leader must demonstrate competence in basic airway skills, use of emergency medications, and the ability to interpret electrocardiograms. The most important skill, however, rests in the ability of the hospitalist to function as an effective team leader with an ability to communicate clearly, concisely, and calmly and demonstrate a capacity to coordinate the activities of other care professionals they might be meeting for the first time.

Hospitalists also add value by actively participating in hospital-wide committees that review the outcomes of rapid response team actions and other emergency cardiac care activities. Only then can they facilitate change in the processes the hospital puts in place for recognizing and responding to the patient with unexpected changes in clinical condition.

LEADERSHIP

Medical staff or hospital-wide committees

In recent years, The Joint Commission, the National Quality Forum, and the Centers for Medicare and Medicaid (CMS) have built standards and endorsed safe practices around medical staff and medical center leadership’s responsibility and accountability for the safety and quality within their organizations. This focus provides the hospitalist with important opportunities to take leadership roles on medical staff and hospital committees, working groups, and task forces that focus on safety and quality. As physicians who concentrate wholly on hospital-based care, no group is better positioned to influence outcomes than hospitalists. They should work as solution seekers for the best ways to standardize handoffs, design the electronic health record, improve infection control practices, create accountabilities for the CMS Core Measures, and oversee team building and rapid response teams throughout the entire organization. Failure to engage in these efforts represents significant lost opportunities.

Safety culture

As hospitalists are clearly some of the most visible physicians in any organization, they bear a unique role and responsibility for promoting a robust “safety culture” within the organization and specifically in the units where they focus their practice. The Agency for Healthcare Research and Quality identifies key features of a “safety culture” that includes the willingness of hospital staff to openly communicate concerns about patient care on their units. Units where staff express comfort when questioning physicians or other authority figures when they disagree or have concerns are considered units with a “safer” culture. Other positive attributes of “safe” units are those in which mistakes are openly discussed and those discussions focus on “systems” issues instead of blaming specific individuals.

The difference between the culture of one unit to that of another in the same hospital often correlates with the difference between the middle management communication styles of the associated units. Units with nursing managers and physician leaders who focus on open, honest, effective, and nonpunitive communication related to adverse patient events score higher on safety culture surveys than others. In addition, there are data to suggest that those units with positive safety culture survey results have a lower incidence of adverse patient events when compared to units with less positive surveys. To that end, the hospitalist is uniquely situated to foster a culture of curiosity, inquiry, and appropriate challenging of authority through role modeling and mentoring. By setting positive examples for other physicians and staff, the hospitalist can lead in the efforts to ensure all of the units in which they work strive toward a safe, patient-centered approach to medical care.

MANAGING PATIENT SAFETY EVENTS

THE PRINCIPLED APPROACH

Even when hospitalists do their best to proactively maximize the safety of their patients, unintentional harm still occurs. Importantly, the integrity of the individual physician or institution rests on the response to patient harm as much as it does to prevention. When harm occurs there is a choice to deny, minimize, rationalize, and blame others, including the patient, or to approach each harm event with a commitment to an open inquiry and honest communication following harm—the “principled approach.” The hospitalist is constantly presented with this choice.

It is well recognized that there are a multitude of barriers to honest communication following harm that include fear of litigation, humiliation, lost income, reputational damage, risk to privileges and license, and the uncertainty of outcome. Nonetheless, the “deny and defend” approach to patient harm and the delegation of managing harm to the legal community has arguably not prevented any of those feared outcomes and instead has damaged the reputation of the medical profession and prevented any learning following patient harm events. The “principled approach” to patient harm is arguably the smarter approach because it effectively addresses many of the reasons that patients sue (lack of communication, need for explanation, sense of dishonesty or “hiding something”) and provides a forum for learning and improving patient safety.

The principled approach to patient harm relies heavily on the hospitalist in at least six specific areas: (1) the immediate response, (2) reporting of harm, (3) communication, initially and in follow-up, (4) investigation, (5) identification of process and performance improvement opportunities, and (6) the necessary follow-up. See Table 20-2.

Responding and reporting

Responding to and the reporting of patient safety events are the first step in any principled process to patient harm (Table 20-3).

Reporting triggers the institutional response process, while the health care team responds to the immediate medical needs of the patient. Most hospitals encourage care professionals, especially physicians, to report any patient safety incident to its Safety and/or Risk Management Department. Reports are often made by telephone, handwritten, online, and in person. Hospitals are mandated by the Centers for Medicare and Medicaid and The Joint Commission to provide for a reporting process for patient harm events.

It is incumbent upon the hospitalist to understand and appreciate the reporting process used in any hospital where they work. Importantly, the hospitalist needs to recognize the importance of their role in taking care of the patient when harm occurs and ensuring that neither the patient nor the family is abandoned during this critical time.

Benefits of reporting

The benefits of rapid institutional reporting of patient safety events within the organization provide substantial incentive for all hospitalists to report and encourage reporting harm events. These benefits include (1) the activation of the internal patient safety and risk management processes including a crisis management plan, if indicated, (2) the preservation of data and information, (3) the opportunity to trigger immediate support for patient, family, and care professional, (4) the initiation of a “quality committee” investigation and the “legal privilege” most states afford such investigations, and (5) the establishment of a communication link with the harmed patient and his or her family.

As with documentation in the medical record, when reporting a patient safety event, hospitalists should take care to provide only the necessary factual information to commence an investigation. They should avoid documenting speculation, hasty conclusions with incomplete facts, or “finger-pointing” in the report.

Investigation

As advocates for quality medical care and patient safety, hospitalists possess special skills for participating in the investigation of serious adverse outcomes (see Table 20-4). Patients and families want and deserve the “facts” after a harm event. An appropriate investigatory process is needed to provide them with the necessary information. The hospitalist should commit to participating in any institutional root cause analysis or other investigatory process following a serious harm event. Such investigations should try to avoid the traditional “shame and blame” approach to adverse events and instead focus on system-based issues and identification of possible areas of improvement.

Nonetheless, prior to knowing all the facts, patients and families are still entitled to effective communication in the early aftermath of a harm event.

Communication

Once harm occurs, honest and effective communication helps maintain trust between the patient and family and care professionals. With their easy availability, often 24 hours a day, the hospitalist is uniquely positioned to facilitate such communication. After all, for the patient who has experienced an unexpected outcome, every hour that goes by without effective communication constitutes more harm.

Honest and effective communication after a harmful adverse event is not just the right thing to do, but the smart thing to do as well. Patients and families sue, in large part, because they perceive a lack of transparency, abandonment, or “cover-up.” A transparent process with open lines of communication and disclosure of all pertinent information can mitigate those patient and family perceptions.

Disclosure of any actual error associated with the harm is a multifaceted process that requires careful planning, preparation, and coordination by physicians and hospital administrators. Given its complexity, physicians understandably fear that an inadequate or poorly executed disclosure of medical errors will only serve to frustrate frontline practitioners, ruin the reputation of the organization and individual practitioners involved in the incident, and encourage lawsuits.

Successful adverse event response programs that include “full disclosure” rely heavily upon integration between the clinical departments and hospital risk management. This integration ensures that the various stakeholders are “on board” or at least aware of the plan for communication after adverse events. The stakeholder list must include the medical malpractice insurance carriers for the various parties who might be affected. In order to provide a consistent approach to adverse events for providers, patients, and families, all of the steps involved in the response to harm should be preapproved by all appropriate stakeholders, before implementation.

The process of communication after harm occurs generally falls in three phases: immediate, intermediate, and final or follow-up phase (Table 20-5). The extent of hospitalist involvement in this type of communication will depend upon the relationship of the hospitalist with the patient and his or her family. Long-term close physician-patient relationships are conducive to multiple meetings and discussions wherein the hospitalist will quickly discover that “disclosure” is a process and not an event.

A liaison for the family should be identified and appropriate empathy and assurance of nonabandonment is expressed at each visit. It is critical for the hospitalist to understand the difference between “empathy” and “apology.” Empathy is the understanding or sharing of another person’s emotions and feelings whereas an “apology” is admitting a mistake that caused harm. In medicine, there is an ethical imperative to express empathy when patients suffer harm but apologies should be reserved for situations when it is clear mistakes or errors have caused the harm. Therefore, apologies should only be offered when the facts are clear and agreed upon by the stakeholders with knowledge of the event.

Example of expression of empathy: “I am sorry your pneumonia has progressed to the point where despite our best efforts we now need to put a breathing tube in your windpipe to help you breathe.”

Example of apology: “I am sorry I did not check your abnormally low blood sugar result this morning. If I had seen the result, I could have given you some extra sugar to prevent your seizure this morning.”

PROCESS IMPROVEMENTS

The value in a principled, transparent approach to adverse patient events lies in the ability to learn from mistakes within a rigorous, reflective environment that promotes performance improvement efforts designed to significantly improve the delivery of care. To be effective, hospitalists must play a role in these performance improvement efforts that follow suboptimal care. Patients and families involved in adverse events caused by inappropriate care are intensely interested in learning the ways in which similar events are less likely to occur. Discussing these quality improvement measures with them also helps to maintain the trust and bond between patient and provider.

CONCLUSION

The evidence for abundant opportunities to improve patient care and prevent patient harm has become indisputable. Earlier estimates on the annual number of preventable deaths in hospitals far underestimated the actual number of deaths that could be prevented, especially from health-care-associated infections. Moving forward it has become clear the mantra “no outcome, no income” or “no pay for low-quality performance” is going to apply to a significant number of episodes of care. To that end, hospitalists are ideally situated to positively influence issues in quality of care outcomes through their active engagement on performance improvement efforts, their communication skills, and their ability to lead multidisciplinary efforts.

In addition, hospitalists can take a lead in providing honest, open, and effective communication to patients and families after unexpected adverse event outcomes. This transparent approach can be the catalyst, for the transformation of an organization’s culture for effectively responding to the needs of patients, providers, and the health care system. Adopting a policy and practice of transparency related to harm events represents a major shift in organizational focus and will need the full support of hospitalists to fully implement. This approach will require strong and persistent endorsement by the kind of leadership that hospitalists can provide their organizations. The added value of transparency is found in the opportunity to rapidly learn from, respond to, and modify practices based on harm investigation with these now transparent events.

SUGGESTED READINGS

INTRODUCTION

Achieving better health outcomes for patients and populations requires a focus on continuous quality improvement (QI). While physicians pride themselves on being subject matter experts in their focused area of medical practice, such knowledge alone is insufficient to produce fundamental changes in the delivery of health care. Physicians who practice in complex hospital and health care systems must acquire another kind of knowledge in order to develop and execute change.

W. Edwards Deming, an American statistician and professor who is widely credited with improvement in manufacturing in the United States and Japan, has described this knowledge as a “system of profound knowledge” (Figure 21-1). This knowledge is composed of the following items: appreciation for a system, understanding variation, building knowledge, and the human side of change. These concepts are just beginning to be taught to health care professionals and are essential for anyone who wishes to improve the health care delivery system.

All hospitalists have witnessed changes that did not result in fundamental improvements within their hospital systems: the computerized order set that was successfully implemented but never revised based on prescribers’ feedback, the paper checklist for medication reconciliation that never gets filled out, or the new rounding system that worked for the first few weeks but then failed to become a standard part of practice due to physician variation or lack of commitment. These are all examples of first-order changes—changes that ultimately returned the system to the normal level of performance. In quality improvement work, individuals must strive for second-order changes, which are changes that truly alter the system and result in a higher level of system performance. Such changes impact how work is done, produce visible, positive differences in results relative to historical norms, and have a lasting impact. Although the model for improvement described below may seem simple, it is actually quite demanding when used properly; and the process is essential to both learning and ultimately changing complex systems.

PLAN-DO-STUDY-ACT AS A TOOL FOR QUALITY IMPROVEMENT

The Plan-Do-Study-Act (PDSA) model is a commonly used method in quality improvement. Shewart and Deming described the model many years ago when they studied quality in other industries. This model first appeared in health care when Berwick described how the tools could be applied using an iterative approach to change. Using a “test-and-learn approach” in which a hypothesis is tested, retested, and refined, the PDSA cycle allows for controlled change experiments on a small scale before expansion to a larger system. The four repetitive steps of PDSA—plan, do, study, and act—are carried out until fundamental improvement, which can be exponentially larger than the original hypothesis, takes place (Figure 21-2).

PLAN

During the Plan phase, the team generates broad questions, hypotheses, and a data collection plan. It is critically important during this period to define expectations and assign tasks and accountability to every team member. In the planning phase of the PDSA cycle, it is prudent to invest significant time and develop a well-framed question by reviewing related research and local projects and defining meaningful process and outcome measurements. Broad questions at the outset of a PDSA cycle can include “What are we trying to accomplish?” and “What changes can we make that will result in an improvement?” The ideal data collection tool answers the question: “How will we know that a change is an improvement?” It is also helpful for the team to generate predictions of the answers to questions early on. This aids in framing the plan more completely, to uncover underlying assumptions or biases before any testing, and to enhance learning in the Study phase by providing a baseline point of comparison.

Teams new to QI frequently will struggle with the question, “How do we measure improvement?” Defining discrete process measures is a good starting point when using PDSA. Process measures are used to assess whether the cycle is being carried out as planned. This is in contrast to outcome measures, which are used to track success or failure and focus on the specific outcome that the team is trying to achieve.

DO

The Do phase in PDSA is a period of active implementation. It involves feedback on the new process from end users and rigorous data collection. An overarching goal of this phase is to capture and document not only compliance with the new process, but also deviations, defects, or barriers in the process. There are always aspects of quality improvement projects that do not go as planned, and flexibility and open-mindedness are critical to maximize learning from improvement. The quality of the Do phase is intimately related to the quality of the Plan phase. A pitfall for many novice QI teams is to give in to the temptation to jump straight to implementing change without spending a significant amount of time planning. A poorly conceptualized improvement plan, an absence of a sound data collection model, or unclear accountabilities can have adverse effects on the implementation or “do” phase of a new initiative.

STUDY

Analysis of available process and outcome metrics and a qualitative appraisal of the process are the key activities in the Study phase. Time should be set aside to perform a critical review of the data collected and compare it to historical data (when available) and baseline predictions. Close attention should be paid to possible defects in any element of the process, including the data collection plan. If such issues are uncovered, the team may need to revise the initial data collection tools and overall plan. Thoughtful review of all trials, even those that were clearly unsuccessful based on metrics, is a critical and valuable process for the team. In fact, the “failures” in a PDSA cycle can yield unanticipated and improved directions. As the Study phase progresses, time should be spent considering if a follow-up PDSA cycle is planned and exactly what elements to include in that cycle.

ACT

The final component in a PDSA cycle is Act. The team should convene for a feedback and action planning session. Frontline workers in the system that is being changed should be included for honest input. A team approach rather than a “top-down” approach facilitates an open review of successes and failures. An action plan that encompasses lessons learned in the first three steps should then be put into motion. During this stage decisions are made about repeating certain test cycles after improvements are made or “spinning off” new test cycles based on the original one.

RAPID CYCLE, CONTINUOUS, AND SEQUENTIAL PDSA

In its most basic form, the PDSA model described above can be applied to change a single process. However, teams in health care often confront problems that require multiple changes, in parallel or succession, in order for improvement to happen. Caution is advised when initiating several PDSA cycles simultaneously, especially if there are significantly different data collection plans or if the team is inexperienced in QI methods. An alternative is a sequential PDSA model in which one PDSA cycle feeds into the next. This approach, in which teams continually change and refine their processes based on data evaluation and feedback, is called “continuous quality improvement.” Experienced QI teams strive to utilize this approach. Rapid cycle PDSA is a continuous QI process that lends itself well to projects that are focused on relatively small-scale changes. It is typically used by seasoned QI teams who are familiar with the PDSA model and who wish to implement rapid change.

AN EXAMPLE OF PDSA IN ACTION

To illustrate the PDSA model for improvement, a real QI project is presented here from start to finish. A hospitalist group sought to implement a new discharge planning toolkit aimed at improving transitions in care through risk assessment at the time of hospital admission. A QI team was formed with representatives from health care professionals involved in the discharge planning process. While their ultimate goal was to reduce unplanned readmissions, their first team goal involved creating a new process to coordinate and request risk-specific interventions from other teams (eg, nurse educators, pharmacists, a nurse for postdischarge follow-up phone calls) for patients deemed “high risk for hospital re-admission or transition in care problems” by a screening tool. In preparation for the project, the QI team also performed a stakeholder analysis, which is a tool that QI teams can use to identify all of the individuals and groups with a “stake” in the process being discussed.

CYCLE 1

PLAN

The initial PDSA cycle involved piloting a readmission risk screening tool. A weekly meeting was convened that included representative users of the tool and assigned specific responsibilities and tasks with due dates to each team member. At baseline, the biggest barriers to overcome were the perception that the new tool was extra work, introducing a paper-based tool in a largely electronic health care environment, and lack of a tight infrastructure tying the requests to existing risk-specific interventions. Based on these concerns, the team reduced the number of interventions on the initial tool. A data collection plan was started and included both quantitative process metrics (eg, compliance rates with the tool, frequency of risk factors identified on the tool) and qualitative data from the users of the tool.

DO

The new tool was piloted for 2 weeks, during which time data was collected and feedback was solicited from the frontline team.

STUDY

After 2 weeks, the data showed that overall compliance with the tool was moderately high, but that two risk factors, health literacy and depression, had unexpectedly low percentages. On further inquiry, members of the team admitted that when they performed the risk screen, they paused on those two questions and frequently left them blank, concerned that it might take too much time during the admission process and frustrate new users of the tool.

ACT

The team decided to make another edit to the tool before the second PDSA cycle. The health literacy and depression screening questions were removed based on feedback, with a plan to reintroduce them when the tool was more embedded in the hospital admission workflow.

CYCLE 2

PLAN

The second PDSA cycle focused on follow-up data collection with the health literacy and depression screening questions removed from the tool, to test the theory that this would improve compliance. The data collection plan was to track overall compliance with the tool for a 2-week period. In order to isolate any improvement as a result of this one small change, no other changes were made during this time.

DO

The new version of the tool was implemented.

STUDY

Compliance rates significantly increased from moderately high to very high, and the risk factor screening data remained unchanged. Qualitative feedback from frontline users was that the risk screening process was more streamlined and acceptable.

ACT

A brief but successful cycle 2 ended with a plan to add an intervention checklist to the tool in the next phase.

CYCLE 3

PLAN

The goal of cycle 3 was to associate risk-specific interventions (education, follow-up phone calls, and social work interventions) with a patient’s individual risk factor profile. To meet this goal, the team implemented a new version of the tool that included the risk-specific intervention requests and tracked request type and volume. A 2-week cycle was planned with continued weekly meetings during this time.

DO

The team implemented a tool that allowed for interventions to be requested at the time of risk factor screening.

STUDY

After 2 weeks, the data showed stable high compliance with the form, stable risk factor data, but very low utilization of intervention requests. At feedback meetings, frontline users stated that at the time of admission, they were not ready to place a request for an intervention. They felt that intervention requests should be discussed in a multidisciplinary team on a follow-up hospital day when more information was available.

ACT

Discharge planners on the team suggested that the intervention request process be integrated into daily discharge planning rounds, during which the entire patient care team (physician, nurse practitioner, registered nurse, discharge planners, patient service representative, and social worker) discussed each patient on the service. A nurse practitioner and patient service representative drafted paper forms that could be used to communicate requests for each of the interventions to the appropriate personnel and to document completion of the task. The next phase would trial this new process.

CYCLE 4

PLAN

Cycle 4 was focused on implementing and studying the new discharge rounds process to request risk-specific interventions. The frequency of intervention requests in each category was added to the existing process metrics. Since this was a more substantial change than before and involved more than just one team of frontline users, a 4-week cycle duration was chosen.

DO

Clinicians continued to screen patients using the risk screening tool, intervention request forms were kept on hand during discharge rounds, and the patient service representative and discharge planners prompted the teams to request interventions based on patient risk factors. The requests were forwarded to the appropriate personnel (registered nurse, pharmacist, nurse educator), who then documented completion of the intervention on the form.

STUDY

Compliance rates and risk factor data remained steady, but there was a significant increase in intervention requests in all categories. However, documentation of completion of the intervention was low. It was determined that the documentation requirements were unfamiliar to the intervention teams, which was an oversight.

ACT

For the next cycle further improvements in documentation of intervention completion and a reintroduction of the health literacy and depression screening questions was planned.

LESSONS LEARNED

This example illustrates several important points for successful use of the PDSA model. First, the engagement and involvement of the end users of new QI tools and processes is critical to the success of any improvement project. These users are experts in the process who often know what should be tested next, and perfect champions when changes are disseminated on a larger scale. While it may be impossible to address or fix every problem that they identify, hearing their input, implementing changes based on their suggestions, and giving praise for their involvement and patience is an important skill for leaders of QI. Second, flexibility and creative thinking, skills that are used frequently in clinical care, are also essential in QI. In the case study, several barriers were identified such as: concerns about paper forms, perception that certain risk factors would halt the risk screening process, and lack of infrastructure around the systematic documentation of interventions. As these barriers became apparent, the team remained flexible and changed a part of the new process without compromising the integrity and team goals of the project.

ALTERNATIVE MODELS OF QUALITY IMPROVEMENT

In addition to the PDSA model described above, there are other frameworks that have been used to design and execute quality improvement projects. Adopting one specific framework (as opposed to adopting several) allows an organization to learn a common language and approach to improvement. Six Sigma and Lean are two common frameworks that will be briefly described.

Six Sigma was developed by Motorola in the mid-1980s and is focused on reducing variations in a process. Six Sigma is a popular performance improvement methodology which uses a five-phase approach to problem solving, called DMAIC (Define, Measure, Analyze, Improve, and Control). This framework guides users to define their QI goals, measure the current process, analyze root causes of the quality problem, improve the process on the basis of the previous steps, and finally control the process to ensure that variances are corrected before they result in defects and the new process becomes standard work.

Lean manufacturing (or just “Lean”) was adapted from the Toyota Production Systems and is focused on continuously reducing waste in operations and enhancing the value proposition to customers. The Lean approach is based on a few key principles: defining the problem from the customer perspective, identifying the activities required to provide the customer with a product or service, producing the products or services only when needed by customers, and pursuing perfection in the process.

CONCLUSION

Plan-Do-Study-Act has remained a fundamental tool for continuous quality improvement. Once comfortable applying this iterative approach, hospitalists can affect both small- and large-scale changes in their health care systems. Other QI frameworks that hospitalists should be aware of include Six Sigma and Lean methodologies. Adopting one specific framework (as opposed to adopting several) allows hospitalists and organizations to learn a common language and approach to improvement.

SUGGESTED READINGS

INTRODUCTION

The practice of medicine is at heart an exercise of collecting, filtering, summarizing, managing, analyzing, and acting upon information. This information comes directly from the patient’s narrative history, but also from family and caretakers, and other providers. It is also derived from diagnostic interventions, including the physical examination, laboratory tests, radiologic exams, and procedures. Combined with reference knowledge about physiology, pathology, pharmacology, and other basic science disciplines, the physician makes an expert assessment of the patient’s conditions and risks, and then recommends an action plan. Information about this plan must be communicated and coordinated with a larger team and with the patient and their family, executed, and then information about how the patient responds fed back, in order to make adjustments over time. If this flow of information is compromised or hampered at any point in this cycle, then the potential for quality and safety problems emerges. Given this intense information-rich environment that the clinician must navigate, especially in the inpatient setting, it is clear that the judicious application of information technology (IT) can greatly empower the hospitalist in providing high quality and safe patient care; and conversely, that injudicious application of IT can promote errors and adverse outcomes.

Information technologies that impact patient safety and quality of care can be grouped into three major categories. First, there are the interventions that impact care as it is delivered in real time—this class is generally called decision support because it involves clinicians while they are making diagnostic and therapeutic decisions. The second class of information technologies, broadly known as surveillance, monitors the immediate downstream care processes to detect anomalies and unintended consequences so that effective corrective action may be taken quickly. The last general category of IT for safety and quality is data mining, or retrospective analysis of large repositories of data, such as patient registries, electronic health records (EHRs), and administrative databases in order to detect meaningful patterns and signals that may help inform ways to improve one or more health care delivery processes. Data mining overlaps with classical epidemiological health services outcomes research.

DECISION SUPPORT

As defined above, decision support is any type of information system that intends to direct, guide, or alter medical decision making as it occurs in real time. This may occur via passive delivery of knowledge, such as quick access to online digital references, drug compendia, clinical calculators, or differential diagnosis tools. In this case, the user must voluntarily choose to activate the service. This type of decision support is usually well received by busy clinicians, because the clinician is motivated to get a question answered. However, passive decision support does not address latent information needs, or knowledge deficits unknown to the clinician.

Decision support may also occur via active knowledge delivery, such as alerts to avoid unsafe or undesired behavior, or reminders to promote desired behavior; the service is activated automatically. Usually, the intended behavior is evidence based, such as avoiding drug combinations that have been shown to result in adverse effects; but it can also be policy driven, such as to promote some medications over others based on formulary or insurance criteria. As active decision support is often interruptive, clinician acceptance of this information is variable, depending upon the perceived usefulness of the information provided and the manner in which it is displayed.

There are certain decision support systems which fall somewhere between active and passive, which facilitate workflow. Examples include messaging systems such as signout applications and secure e-mail or text paging, electronic medication reconciliation applications, and results management programs. Also in this category is the organized presentation, or summarization, of patient data that lowers cognitive burden.

The ideal decision support intervention simultaneously facilitates the desired workflow(s) while impeding the undesired workflow(s). In other words, the most effective decision support interventions make it “easy” to do the right thing, and “hard” to do the wrong thing. For example, compare two ways to implement decision support for optimal drug dosing. The first, more common approach is to analyze medication orders after they are entered; compare them to rules that assess patient factors such as age, gender, and comorbidities such as renal dysfunction; and then display a series of corrective alerts. The second approach does as much of the patient-specific calculations as possible up front, so that only the most reasonable medication alternatives for a given indication are offered in the first place, with default dose and frequency precalculated to match the patient’s condition. Only the prescriber who chooses to override the defaults is interrupted to provide an override reason. Of course, the more sophisticated consultative approach to delivering decision support requires more data in computable form about a patient, as well as more complicated and nuanced rules, than the typical critical approach.

Evidence suggests that decision support can successfully influence provider behavior, improve process measures, increase quality of care, and reduce errors and adverse events. However, the way decision support is implemented can dramatically affect effectiveness. For example, when influenza vaccination reminders were first implemented at one inpatient site, the effect was minimal; but when the alert was changed to a complete prewritten order, and the default was set to “accept” instead of to “decline,” the inpatient vaccination rate increased from 1% to 51%.

More problematic is the growing recognition that poorly implemented systems can have unintended adverse consequences on health care providers, to the point of inciting clinician revolt, such as occurred at Cedars-Sinai Medical Center in Los Angeles in 2003. Worse yet, computerized provider order entry, coupled with decision support, can potentially promote errors. Therefore, the behavior of such systems must be continually scrutinized, and the information they provide must never be accepted blindly.

SURVEILLANCE

Surveillance IT is analogous to secondary prevention or care—it is meant to detect complications of care early so that the consequences can be prevented or ameliorated. The most prevalent example, which is still relatively rarely implemented in practice, is adverse event detection. This requires an electronic monitoring system that oversees all digital transactions in a health information system, such as new orders, new lab results, and new patient encounter records; a repository of rules that define potential events and the actions that should be taken; and a variety of effector systems to carryout the actions, such as texting or e-mail alerting. Collectively, these components form what is commonly known as an event engine. As an example, the monitoring system registers all new lab results, including an individual patient’s falling platelet level. A rule in the repository defines a clinically significant rapid rate of decline (eg, an absolute drop of 50,000 or a relative drop of 30% over 2 days, or of 75,000 or 50% over 3 days) in the right clinical context (the patient has a current active order for a heparin-containing medication), to generate a response (alert the responsible clinician the next time she logs into the system, or text page a backup clinician if this does not occur within 24 hours). It is clear from this example that the effectiveness of this surveillance system is only as good as the breadth of events that can be monitored, the granularity with which rules can be authored to define clinically significant events, and the breadth of interventions available for actions.

Surveillance systems can collect and analyze quality-related data, as well as safety data. Such systems, sometimes called profiling or detailing systems, have been long utilized by pharmaceutical companies to direct and tailor marketing efforts, but they can also be used to track how often a hospitalist utilizes nonformulary medications; or how well a provider is achieving quality of care metric goals, such as percentage of their patients who have a Foley catheter for over 48 hours; or rates of resource utilization, such as magnetic resonance imaging for headaches. This information can be shared with just the relevant provider, or with an entire practice, either deidentified or not. The most sophisticated profiling systems present the data in a quality dashboard that the provider can query dynamically in real time and link the data to relevant actions, such as e-mailing patients, flagging them for callback appointments, or automatically referring patients to a disease management program.

In addition to patient-specific and provider-specific event engines, surveillance systems have also been developed to work at the population level. For example, there are monitoring systems that track aggregated data about visits to regional emergency rooms, including chief complaints, to detect early signals of disease outbreaks such as from influenza or bioterrorism.

DATA MINING

The final class of IT that can be brought to bear on quality and safety is the retrospective analysis of large datasets to look for trends and patterns and their relationship or association with significant events, interventions, or behaviors. This traditionally has been called health services research, and can focus on either health care outcomes (mortality, morbidity, readmissions, adverse event rates) or their process-based proxies (frequency of deep vein thrombosis [DVT] prophylaxis measures, rate of compliance with recommended guidelines, or proportion of completed discharge summaries within 24 hours). Both are valid indicators of quality and safety effectiveness, though hard outcomes are often preferred, but usually more difficult to measure and influence. For example, measuring the effect of an intervention on incidence of DVT would be ideal, but in practice would require significant manual data collection by chart abstraction, and may be too infrequent an event in the time window allowed for study to make statistically sound conclusions. Instead, measuring how the intervention impacts the number of orders for subcutaneous heparin, especially where such orders are placed via computer-based provider order entry (CPOE), is much easier and a more common event.

The newest direction this type of research has taken is the combination of large datasets from different disciplines, to look for new and sometimes unanticipated or counterintuitive associations. Because of the increasing likelihood of chance alone being responsible for observing such relationships between data when multiple statistical tests are performed with the same data, this type of knowledge discovery requires large collections of data, runs the risk of uncovering statistically significant but clinically irrelevant patterns, and should always be considered hypothesis generating rather than confirming or refuting. A simplistic example is that if one were to measure how frequently lung cancer patients carry matches compared to patients without lung cancer, one might be tempted to conclude that carrying matches is a very dangerous activity. Another example is that since test results are typically defined to be in the normal range when they lie within the 95th percentile of results from a healthy population, then a battery panel of 20 tests will have at least one false positive result almost two-third of the time.

MEANINGFUL USE

In the United States, a recent potent driver of health IT efforts has been Meaningful Use (MU). Through the Health Information Technology for Economic and Clinical Health Act, up to $27 billion (or over $60,000 per clinician) over 10 years is set aside to promote not just the adoption, but the “meaningful use” of Electronic Health Records in order to achieve improved patient outcomes. This is being pursued incrementally via three stages enacted over time.

• Stage 1 (2011-2013): Capture health information electronically in a structured format; use that information to track key clinical conditions and coordinate care; implement clinical decision support tools; use EHRs to engage patients and families and report clinical quality measures and public health information.

• Stage 2 (2014-2016): Encourage the use of health IT for continuous quality improvement at the point of care and the exchange of information in the most structured format possible.

• Stage 3 (starting 2017): Promote improvements in quality, safety and efficiency leading to improved health outcomes by focusing on decision support, patient access to self-management tools, access to comprehensive patient data and improving population health.

Compliance with the initial stages is incentivized with bonus payments from Medicare and Medicaid to both eligible providers and to hospitals. These payments decrease over time and eventually disappear, to be replaced by payment penalties starting in 2015 for noncompliant providers and hospitals.

There are growing data to support the claim that MU Stages 1 and 2 have had a major impact on the use of EHRs and clinical decision support (CDS) in the United States. For example, a longitudinal study of 493 nonfederal hospital IT budgets from 2009 to 2011 found increases in the percentage of hospital annual operating budgets allocated to IT in the years leading up to these federal incentives. Adoption of EHRs by US hospitals has increased dramatically since the MU program was implemented. In 2011, 833 hospitals and 57,652 professionals attested for MU. The following year, an additional 1726 hospitals and 132,395 professionals attested. According to CMS data presented in April 2013, over 77% of eligible hospitals and approximately 53% of eligible providers qualified to earn the bonus. The data also indicate major increases in CDS use compared to 2008 baseline.

On the other hand, data suggesting that the MU program actually has improved health IT is mixed. For example, 60% of hospitals responding to a survey reported that their EHRs did not contain all features essential for high-quality care. And data that MU has positively impacted patients are indirect and inconclusive.

THE HOSPITALIST’S ROLE

There is great opportunity for clinicians such as hospitalists, who are often experts in workflow and systems thinking (whether by formal training or simply by experience), to help guide implementations of decision support interventions within their practice sites in order to increase the chance of success. Even after successful implementation of information technologies, there is ongoing need for clinicians to provide feedback about what works, what does not, and what could be done to improve the system. A higher level of involvement of hospitalists is to provide subject matter expertise to tweak rules and author new ones to make the decision support more specific, relevant, and effective. This never-ending work to keep the content of rules in line with ever changing knowledge has emerged as a new field of its own, called knowledge management.

SUGGESTED READINGS