Chapter 8. Transition and Handoff Errors

An 83-year-old man with a history of chronic obstructive pulmonary disease (COPD), gastroesophageal reflux disease, and paroxysmal atrial fibrillation with sick sinus syndrome was admitted to the cardiology service of a teaching hospital for initiation of an antiarrhythmic medication and placement of a permanent pacemaker.

The patient underwent pacemaker placement via the left subclavian vein at 2:30 pm. A routine postoperative single-view radiograph was taken and showed no pneumothorax. The patient was sent to the recovery unit for overnight monitoring. At 5:00 pm, the patient stated he was short of breath and requested his COPD inhaler. He also complained of new left-sided back pain. The nurse found that his oxygenation had dropped from 95% to 88%. Supplemental oxygen was started and the nurse asked the covering physician to see the patient. The patient was on the nurse practitioner (NP)-run non-house staff service; however, the on-call intern provides coverage for patients after the NPs leave for the day.

The intern, who had never met the patient before, examined him and found him already feeling better and with improved oxygenation after receiving the supplemental oxygen. The nurse suggested a stat x-ray be done in light of the recent surgery. The intern concurred and the portable x-ray was completed within 30 minutes. About an hour later, the nurse wondered about the x-ray and asked the covering intern if he had seen it. The intern stated that he was signing out the x-ray to the night float resident, who was coming on duty at 8:00 pm.

Meanwhile, the patient continued to feel well except for mild back pain. The nurse gave him analgesics and continued to monitor his heart rate and respirations. At 10:00 pm, the nurse still hadn't heard anything about the x-ray, so she called the night float resident. The night float had been busy with an emergency but promised to look at the x-ray and advise the nurse if there was any problem. Finally at midnight, the evening nurse signed out to the night shift nurse, mentioning the patient's symptoms and noting that the night float intern had not called with any bad news.

The next morning, the radiologist read the x-ray performed at 6:00 pm and notified the NP that it showed a large left pneumothorax. A chest tube was placed at 2:30 pm, nearly a full day after the x-ray was performed. Luckily, the patient suffered no long-lasting harm from the delay.1

In a perfect world, patients would stay in one place and be cared for by a single set of doctors and nurses. But, come to think of it, who would want such a world? Patients get sick, and then get better. Doctors and nurses work shifts, and then go home. Residents graduate from their programs and enter practice. So handoffs—the process of transferring primary authority and responsibility for clinical care from one departing caregiver to another incoming one2—and transitions are facts of medical life.

As we all learned when we played the game of “telephone” as kids, every handoff and transition comes with the potential for a “voltage drop” in information. (It also carries the possibility that a new set of eyes or circumstances will lead to a clinical benefit, but our purpose here is not to explore that optimistic scenario.) In fact, handoff and transitional errors are among the most common and consequential errors in healthcare. Despite this, these mistakes received little attention until recently, in part because, by their very nature, they tend to fall between the cracks of professional silos. As we have come to recognize the frequency and impact of handoff and transition errors, we are beginning to learn how to mitigate the harm that often accompanies them.

Healthcare is chock-full of two kinds of transitions and handoffs.3 The first are patient related, as a patient moves from place to place within the healthcare system, either within the same building or from one location to another (Table 8-1). The second are provider related, which occur even when patients are stationary (Table 8-2).

Both kinds of handoffs are fraught with hazards. For example, one study found that 12% of patients experienced preventable adverse events after hospital discharge, most commonly medication errors (Chapter 4).4 Part of the problem is that nearly half of all discharged patients have test results that are pending at discharge, and many of them (more than half in one study) fall through the cracks.5 In another study, researchers found that being covered, principally at night, by a different physician was a far better predictor of hospital complications and errors than was the severity of the patient's illness.6 The same researchers devised a standardized computerized sign-out form and the error rate fell by a factor of 3.7

Even if we grant that some transitions are necessary, one might reasonably ask whether healthcare needs to have quite so many of them. The answer is probably yes. Research has demonstrated that patients do worse when nurses work shifts longer than 12 hours, and that intensive care unit (ICU) residents make fewer errors when they work shifts averaging 16 hours instead of the traditional 30–36 hours (Chapter 16).8,9 Unfortunately, in a 24/7 hospital, implementing these shift limits automatically generates handoffs (Figure 8-1), such as many of the ones in this case (Figure 8-2). Other handoffs emerge when patients receive appropriately specialized care.

Although some might wistfully long for the day when the family doctor saw the patient in the office, the emergency room, the hospital, and the operating and delivery room, most patients now prefer the additional expertise, training, and availability of specialists in sites of care (i.e., emergency department, ICU, and, increasingly, the hospital10), procedures (delivering a baby), or diseases (heart attack or stroke). As these specialists become involved in a patient's care, they create transitions and the need for accurate information transfer. So too does a patient's clinical need to escalate the level of care (such as transitioning from hospital floor to step-down unit) or the economic realities that often drive de-escalation (hospital to skilled nursing facility).

The presence of all these handoffs and transitions makes it critical to consider how information is passed between providers and places. Catalyzed in part by the mandated reduction in resident work hours in the United States that began in 2003 (Chapter 16), there has been far more attention paid to handoffs in recent years. In my own hospital, for example, the number of handoffs by internal medicine residents rose by 40% after duty-hours limits were implemented.11

There is increased pressure coming from the policy arena as well. In 2006, the Joint Commission issued National Patient Safety Goal 2E (Appendix IV), which required healthcare organizations to “implement a standardized approach to handoff communications including an opportunity to ask and respond to questions.” Spurred on by studies demonstrating staggeringly high 30-day readmission rates in Medicare patients (20% overall, nearly 30% in patients with heart failure),12 in 2012 Medicare began penalizing hospitals with high readmission rates.13 All of this attention has catalyzed research on handoffs and transitions, giving us a deeper understanding of best practices, which have both structural and interpersonal components.

This chapter will provide a general overview of best practices for person-to-person handoffs and patient transitions, and then focus on one particularly risky transition: hospital discharge. Additional information about handoffs and transitions can be found in the discussion of medication reconciliation (Chapter 4), information technology tools (Chapter 13), and resident duty hours (Chapter 16).

Like many other areas of patient safety, the search for best practices in handoffs has led us to examine how other industries and organizations move information around. This search has revealed several elements of effective handoffs: an information system, a predictable and standardized structure, and robust interpersonal communication. In one particularly memorable example, physicians at London's Great Ormond Street Children's Hospital studied Formula 1 motor racing, particularly the pit-stop crews (who switch out a car's tires, gas up the car, clean the vents, and send the car screeching back onto the track—all in seven seconds), for lessons in how to do effective handoffs.14 The differences in the approaches of the pit-stop crews and the surgical teams were striking. Inspired, Great Ormond hired a human factors expert (Chapter 7) and reengineered the way its teams performed their postoperative handoffs (Table 8-3). The new model resulted in a significant decrease in handoff errors.15,16

The Joint Commission's expectations, expressed in its 2006 National Patient Safety Goal, include interactive communications, up-to-date and accurate information, limited interruptions, a process for verification, and an opportunity to review any relevant historical data. Vidyarthi and colleagues developed the mnemonic “ANTICipate” to help structure written sign-outs (Table 8-4). For example, in the case that began the chapter, listing “Tasks” in the form of “if, then” statements might have decreased the ambiguity. The written sign-out might have included: “Check the chest x-ray taken at 6 pm. If clear, call the nurse. If it shows a pneumothorax, call thoracic surgery for possible chest tube.” Contingency plans could have taken the form of: “if the patient is short of breath, try an albuterol inhaler (history of COPD), but also consider pneumothorax (patient had recent line placement).”1

While written sign-outs can take a variety of forms, there is an increasing recognition of the advantages of computerized sign-out systems over traditional index cards. At the University of California, San Francisco (UCSF) Medical Center, we developed a computerized sign-out module (“Synopsis”), which resides within the electronic medical record (Figure 8-3). The template standardizes the content of the sign-out and allows multiple providers to see the same data. It also imports certain information from the remainder of the electronic medical record (including administrative information, vital signs, laboratory studies, medication lists, and resuscitation [“code”] status), while allowing for free-text data entry. As one might expect, systems like this improve the quality of sign-outs and decrease the risk of communication-related errors.7,17,18

Computers are an essential part of the answer, but only part. Even in healthcare systems that enjoy advanced computerized records, person-to-person communication remains both necessary and potentially error prone (one study of internal medicine residents' sign-outs found omissions and mischaracterizations in nearly one in four information transfers19). In civil and military aviation, fields that, like healthcare, depend on the accurate transmittal of information, the so-called phonetic alphabet is often used during voice communications. In this system, a standardized word substitutes for a single letter (such as Alpha for “A,” Bravo for “B,” Charlie for “C,” and so on), which permits one person to clarify a bit of information, such as the spelling of a name, without wasting time thinking of words that might have a common reference. Using this same system in a hospital, “Oscar Romeo” for operating room (OR) would never be confused with “Echo Romeo” for ER.

Read backs are also commonly used in aviation and other industries to prevent improperly received messages. For example, a pilot receiving flight plan clearance over the radio from Air Traffic Control always reads it back. In healthcare, we have long used read-back procedures to verify the identity of blood transfusion recipients, but, remarkably, no tradition of read backs for other transfers of risky information existed until the process was mandated by the Joint Commission in 2004 (Appendix IV). It is one measure of our inattention to patient safety that until very recently a physician would have been far more likely to hear the words, “Let me read your order back to you,” when calling her local Chinese takeout restaurant than when she called the nurses' station of virtually any American hospital.20 In one study of 822 telephone calls for critical laboratory test results, read backs caught 29 errors (3.5%), some of them quite serious.21

The bottom line is that having a standardized process for provider-to-provider sign-out is essential.22,23 Sign-outs should take place in a standard place and at a standard time, and the incoming and outgoing personnel must overlap to the degree needed to transmit the relevant information. Ensuring that sign-outs occur in a quiet room, free of distractions, is critical, as is ensuring that both parties have access to the hospital's information system during the process.24 (On the UCSF medical service, we insist that shift change handoffs occur face-to-face, while end of rotation handoffs, such as when one hospitalist is signing out to another after a 1- to 2-week stint, may take place by telephone, with both parties simultaneously reviewing the electronic health record, sometimes via remote access.)

Beyond these “must do's,” there are many variation on the theme. Some units have instituted a morning briefing to support reliable communication at shift change.25 In the ICU, the use of a daily goal card—a simple device to ensure that physicians, nurses, and patients focused on the same limited set of goals—led to improved communication and shorter ICU lengths of stay.26 A dry erase whiteboard in patients' rooms can promote communication and teamwork among all of the involved providers.27 While all of these interventions seem intuitively sensible and relatively straightforward to implement, by now it will not surprise you to learn that they are actually complex changes that should be viewed through a human factors lens (Chapter 7) to be sure they meet their purpose and don't create unintended consequences.15

Joe Benjamin, a 43-year-old mechanic, came to his local ER with chest pain, and was admitted to the hospital to “rule out MI.” Benjamin's ER physician ordered a chest radiograph in addition to the cardiac enzymes and serial electrocardiograms. Since nobody expected Benjamin's x-ray to show anything exciting, and since there was no checklist or protocol to remind anyone to check it, it was forgotten as soon as it was taken. Twelve hours later, after all of the heart attack tests proved negative, Benjamin was discharged. The discharging hospitalist, unaware of the chest x-ray, didn't think to check its results. Meanwhile, the radiologist had reviewed the x-ray and noticed a small lung nodule. He filled out a report for the hospital chart, and sent a version of it to the primary care physician's office, where it was lost.

Two years later, Joe Benjamin developed a chronic cough and a repeat chest x-ray revealed an obvious lung nodule. The radiologist's report read, “The nodule is markedly enlarged compared to the one seen on the x-ray of October 18, 1999.” That was the first Benjamin's primary physician had heard of the prior x-ray, and it was too late. Eighteen months later, Joe Benjamin was dead of lung cancer.20

In its 2001 report, Crossing the Quality Chasm: A New Health System for the 21st Century, the Institute of Medicine compared the U.S. healthcare delivery system to a railroad whose tracks change gauge every few miles:

Health care is composed of a large set of interacting systems—paramedic, emergency, ambulatory, inpatient, and home health care; testing and imaging laboratories; pharmacies; and so forth—that are connected in loosely coupled but intricate networks of individuals, teams, procedures, regulations, communications, equipment, and devices. These systems function within such diverse and diffuse management, accountability, and information structures that the overall term health system is today a misnomer.28

One potential failure point in all of these interlocking systems is the piece of paper, the basic unit of medical record keeping for centuries. Moving paper across healthcare transitions is inherently risky, as it can easily get lost. Moreover, several people, separated by place and function, often need to see patient data simultaneously, and each may create new data and observations that must be added to the record. Paper is clearly not the ideal medium for this.

Many of the handoff techniques discussed earlier can help here as well—including the use of read backs, standard communication protocols, and checklists (such as for hospital discharge) for patients/families (Table 8-5) and providers (Table 8-6). But part of the problem relates to the absence of integration in most of American medicine. Take, for example, hospital discharge—an inherently risky transition. In the United States, the hospital discharging the patient and the physician receiving the patient are usually not part of the same organization (in most cases, the physician is an independent or small group practitioner). Therefore, the physician's office information system, even if it is computerized, will rarely communicate effectively with the hospital's system, and vice versa. This is a formula for a voltage drop at a most critical time.

Integrated systems of care—in which the same organization owns or runs the doctor's office, the nursing home, and the hospital—have great advantages in this regard. Patients who leave a U.S. Veterans Affairs (VA) hospital for a nursing home generally go to a VA-owned facility; the same is true for patients in the huge Kaiser Permanente system and other large integrated delivery systems. The reason these handoffs generally go better is partly organizational, partly psychological, and partly informational. Organizationally, when a single system is responsible for care on both sides of a specialty or facility gap, they collaborate on smoothing the transition and bear the consequence of any problems. Psychologically, human beings simply communicate better with colleagues than with strangers. Finally, from an informatics standpoint, large integrated systems tend to devote more resources to constructing, maintaining, and improving the computerized linkages so critical to communication.

In the United States, there is considerable policy pressure designed to prod healthcare delivery systems and physicians to higher levels of integration. Upcoming payment changes such as bundling29 (a single payment for an episode of care lasting weeks or months) and the promotion of Accountable Care Organizations30,31 (a delivery system that includes the physicians and facilities [i.e., hospitals, clinics, skilled nursing facilities], is at risk for its cost and quality outcomes, and can receive a bonus for high-value care) are, at their core, efforts to promote integration (Chapter 3). Whether they will succeed remains to be seen, but they are promising initiatives.

Recognizing that most of the American healthcare system will remain fragmented for the foreseeable future, much of the federal effort to computerize U.S. healthcare—backed up by nearly $30 billion in incentive payments—is focused on creating a set of standards and protocols to promote interoperability, even when the computers are manufactured by different vendors and owned by different organizations (Chapter 13).32 The best analogy is banking's automated teller machine (ATM) system, in which interoperability allows customers to make transactions away from their own bank, almost anywhere in the world.

Recently, concerns have been raised about patient privacy and the unauthorized use of medical information, and these concerns have taken on additional force with the prospect of large, interoperable computer systems containing terabytes of often sensitive patient data. These concerns, partly addressed in the United States by the 2003 Health Insurance Portability and Accountability Act (HIPAA), pose a real barrier to developing a national medical database, or even to moving patient information from clinic to hospital and back out again. While abuses are possible (and there have already been some high-profile examples of them), policymakers will have to balance the desirability of strict control over patient data against the value of linking healthcare systems in an increasingly mobile society.33 Both individuals and organizations have experimented with having patients carry their own data (in the form of “smart cards” or even implantable chips) as a way of addressing these privacy concerns.34 Because these systems seem a bit too unreliable and inflexible (one big challenge is that they will need to be updated after every clinical encounter), the more realistic hope will be for a single set of national, or even international, standards that would combine seamless interoperability with robust safeguards that ensure reasonable degrees of privacy.

Before turning to the issues surrounding hospital discharge, it is worth reemphasizing that every transition carries the potential for harm, and all are worth scrutinizing for risk reduction opportunities. For example, transfers within the hospital itself—from the emergency department to the floor, the floor to the ICU, and the OR to the recovery room—are prone to error. A systematic review of these transitions by Ong and Coiera identified a number of key themes and evidence-based improvement strategies (Table 8-7).35

One particularly risky transition occurs when a patient leaves his or her hospital floor for an x-ray or other test. Unlike the patient transferring to the ICU, where there is an established nursing intake protocol and the patient has been identified as high risk, many of these intrahospital transports are done haphazardly, with an abrupt decrease in the capabilities of the monitoring personnel and equipment and ambiguity regarding what actions to take in the event of deterioration.36 One hospital uses a “Ticket to Ride” system, where the ticket—which includes patient identification, stability, and risk information—serves as the communication form between sending and receiving personnel. The transport personnel are responsible for ensuring the nurse completes the ticket, which stays with the patient until return to the home unit. Analysis of over 12,000 transports demonstrated improved transport documentation, higher patient satisfaction with transport, and fewer adverse events.37 A well-designed checklist would probably serve the same function.

While the risks of intrahospital transfers are obvious, it is also worth emphasizing that there are major transition and handoff issues that occur in community practice. One study tracked the information exchange between outpatient providers who cared for a patient following a hospital discharge, and discovered that information from the previous visit was available only 22% of the time.38 This problem is magnified by the sheer volume of information that an outpatient provider needs to manage. In one large practice, the average primary care physician was responsible for following up on 800 chemistry/hematology tests, 40 radiographs, and 12 pathology specimens in a typical week.39 Add to all that the need for consultation, and things look even grimmer. In another study, two-thirds of primary care physicians (and one-third of the specialists) were dissatisfied with the referral process, which isn't a surprise since the majority of specialists reported that they usually received no information from the primary care doctor, and one in four primary care doctors had not heard back from specialists within a month of their referral.40

While the solutions here would seem to rest with better information technology, this is proving harder than it looks. In one study, the VA system created an electronic alert for abnormal imaging results—alerts went to both the receiving physician and the radiologist. Despite what appeared to be a foolproof system, flawed follow-up remained distressingly common.41 Similarly, computerized order entry has been held out as the savior for many errors in transitions and medication reconciliation. Researchers examined 4000 computer-generated prescriptions and found a 12% error rate.42 Clearly, our efforts to solve transitional problems with information technology will need to be accompanied by equal efforts to apply human factors thinking, be alert for unanticipated consequences, and remain attuned to the challenges of trying to change care patterns in complex systems (Chapter 2).

The issues surrounding one distinctive and risky handoff—the end-of-year transition that occurs when residents graduate and turn their practices over to new interns43—will be discussed in Chapter 16.

As discussed earlier, we have seen an abundance of evidence that the hospital discharge transition is one we manage exceptionally poorly. This was highlighted by the seminal study by Jencks et al., which found that 20% of Medicare patients are readmitted within a month of discharge and one-third return within 90 days.12 The authors estimated the cost of preventable readmissions at $17 billion.

Other studies have chronicled a litany of additional postdischarge disasters:

• Nearly 20% of patients experience adverse events within 3 weeks of discharge.4
• Forty percent of patients are discharged with tests pending, and many of these balls are dropped.5
• Fifteen percent of discharged patients have a discrepancy in their medication lists.44
• Astonishingly few discharge summaries reach the primary care physician by the time the patient returns for his or her first postdischarge visit.45

Despite powerful literature that shows that relatively simple interventions—such as postdischarge phone calls or the use of a transitions coach—can lead to striking improvements in postdischarge care and decreased readmission rates,46,47 only a small fraction of hospitals have implemented them. Moreover, few hospitals outside of integrated delivery systems have become electronically linked to their referring physicians' offices.

Recent policy changes such as readmission penalties13 and value-based purchasing (Table 3-5) have turned improving the discharge transition into an imperative for hospitals, primary care doctors, and their respective systems. Luckily, a number of best practices have been identified, with proven beneficial effects on readmission rates and other meaningful outcomes.

In considering these best practices, several themes have emerged. First, hospitals will need to improve their discharge processes so that they can reliably: (a) get a discharge summary in the hands of an outpatient doctor on the day of discharge; (b) ensure that patients receive robust and understandable discharge instructions (not simply a check box on a form) (Figure 8-4 shows a discharge checklist, drawn from the Society of Hospital Medicine's Project BOOST48); and (c) keep in touch with patients via a follow-up phone call (or e-mail or Tweet—whatever gets the job done!). Second, high-risk patients will require something more: a postdischarge clinic visit, perhaps, or a discharge or transitions coach or a high-risk case manager. This isn't rocket science—all these interventions make sense, have manageable costs, and are challenging but not impossible to implement. They simply take institutional will, strong teamwork, and an appreciation of the principles of quality improvement and change management.

Of all of the discharge transition programs, perhaps the most creative has been Project RED (“Re-Engineered Discharge”), developed by a team at Boston University.47 The RED protocol involves giving patients intensive predischarge counseling, a customized discharge document (Figure 8-5), an early postdischarge phone call, and a rapid follow-up appointment. The intervention resulted in a 30% decrease in readmissions and modest cost savings.

One of the most expensive parts of the intervention was the discharge counseling: an RN spent nearly 90 minutes with each patient and family, at significant cost. To address this issue, the BU investigators developed an animated conversational character, nicknamed “Louise.” Louise is preprogrammed with individual patient information, which she combines with answers to disease-specific and general questions. She—a pleasant-looking middle-aged woman of purposely ambiguous race and ethnicity—appears on a computer screen wheeled to the patient's bedside (Figure 8-6). Speaking to the patient in a relatively fluent but unmistakable computer voice (with synchronized animation), she might say: The doctors tell me that your primary condition is atrial fibrillation … it usually causes a rapid heart beat. Do you have any questions about this condition? Patients answer Louise's questions by picking choices off a touch-screen interface. She then responds according to her algorithm.

In a pilot study, Louise's performance was equivalent to that of the human discharge coach, while saving nearly $150 per encounter.49 More interestingly, not only did most patients find Louise to be personable and responsive, but also three out of four preferred interacting with “her” to receiving discharge instructions from a live physician or nurse.

Let's be clear. While Louise can calmly walk a patient down an algorithmic path, answer simple questions, and test basic understanding, dealing with complex situations, managing ambiguity, and showing empathy are not her forte. Nevertheless, Louise's success indicates that certain functions that we call “nursing” can probably be accomplished by other people or technologies, at a lower cost (the same is undoubtedly true of certain doctoring functions as well, I hasten to add). Ultimately, this kind of creative thinking will be needed if we are to meet the challenges of providing high-quality, safe, reliable, and satisfying care at an affordable cost.

• Errors at the time of transitions (also known as handoff errors) are among the most common errors in healthcare.
• Handoffs can be site-to-site (e.g., hospital to skilled nursing faculty) or person-to-person (e.g., one physician signing out to another).
• For both kinds of handoffs, solutions will involve both improved information systems and standardized protocols.
• Handoffs should occur at designated times and without distraction, cover likely scenarios, include “if/then” statements, and utilize read backs and a phonetic alphabet.
• It will be vital to have computer systems that “talk to each other” (“interoperability”); creating these systems will require that we address legitimate concerns about patient privacy.
• Increasing pressure to improve transitions is coming from a variety of policies (public reporting and financial penalties for high rates of readmissions). Luckily, several comprehensive programs have resulted in markedly decreased readmission rates, and practices have been identified that appear to improve other transitions and handoffs as well.