Dropping the Baton: A Qualitative Analysis of Failures During the Transition From Emergency Department to Inpatient Care

Background and Importance 

Transfers of care among providers have been identified as a major source of medical error,1, 2, 3 yet the contribution of transfers to error remains unclear. To reduce errors, a more complete understanding of the processes involved in transferring patient care is needed. Transferring a patient from one care provider to another requires, at the minimum, communication of clinical information. However, a care transfer is much more than 1-way communication. A well-conducted transfer results in seamless continuation of care and is transparent to participating physicians, patients, and staff. Transfers also play an important role in teaching, promotion of team cohesion, emotional support, socialization, maintenance of social order, and error detection.4, 5, 6, 7, 8, 9

Most research on hospital transfers has focused on transfers within a specialty, such as resident-to-resident end-of-shift sign-out.10, 11, 12, 13 The transfer from emergency department (ED) to admitting physicians is little studied,4, 9, 12, 13, 14, 15 yet there are several theoretical reasons it may be of particularly high risk. This transfer must span changes in 3 domains—provider, department, and physical location—which may not occur simultaneously. Cultural, linguistic, and social differences between emergency and internal medicine physicians may increase potential for conflict or misunderstanding.16, 17 Often, uncertainty about diagnosis and treatment is high, yet this uncertainty may not be appreciated because of clinical inertia, cognitive biases, face-saving concerns, or a need to “prove” the patient requires admission.4, 9, 18 Results of tests and studies are frequently still pending, creating opportunities for missed follow-up. ED-floor transfers take place in a setting that can be chaotic, crowded, and rife with distractions.19 Finally, the ED-floor transfer occurs early in the hospital course, when patients may be least stable and thus most vulnerable to effects of failed transfers.

Goals of This Investigation 

A better understanding of failed transfers may help to improve patient safety during this critical point of hospitalization. Accordingly, we aimed to identify vulnerabilities in the ED-floor transfer process through qualitative analysis of failures reported by emergency and internal medicine physicians.

Study Design 

We designed a cross-sectional survey study that was pilot tested for clarity and content by chief residents in emergency medicine and internal medicine (see Appendix E1 and E2, available online at http://www.annemergmed.com). Self-administered, anonymous questionnaires were sent by e-mail and distributed at conferences 3 times in March 2007. A lottery for one of 3 $50 Amazon.com gift certificates was a financial incentive for participation. The Human Investigation Committee approved the study and granted a waiver of signed informed consent to preserve anonymity of respondents. Return of the survey was considered consent to participate.

Setting 

The study was conducted at a 944-bed urban, academic medical center with both emergency medicine and internal medicine residency training programs. The hospital uses a computerized physician order entry system, but ED orders, notes and vital signs are not readily visible to receiving physicians in the electronic medical record. ED house staff and physician assistants provide oral sign-outs about new admissions; no standardized template was in use at the time of the study. Medicine patients are divided approximately equally between a teaching service and a nonteaching service. Sign-out is received by house staff for patients admitted to the teaching service and by attending hospitalists for patients admitted to the nonteaching service.

Selection of Participants 

We sent the survey to all medical staff who could be involved in providing or receiving patient transfers from the ED to internal medicine. This included house staff, physician assistants, and attending hospitalists.

Outcome Measures 

The survey included a question specifically about adverse events, using the definition established by the Institute of Medicine20: “Has a patient of yours ever experienced (or almost experienced) an adverse event because of inadequate communication between the ED and admitting physician? An adverse event is an injury caused by medical management rather than the underlying condition of the patient. You may include ‘near misses,' where the problem was recognized in time to prevent injury to the patient.” If the answer to that question was yes, respondents were then asked to “briefly describe the adverse event or near miss, omitting the patient's name.” More than 1 event could be described. This report describes the results of a qualitative analysis of the responses to this open-ended question.

Primary Data Analysis 

We used qualitative data analysis methods,21 followed by a review of literature in communication theory, human factors engineering, and cognitive psychology to identify and categorize the processes involved in transferring a patient from the ED to the inpatient unit. The study team included emergency physicians (R.G.K., T.M., J.D.S.), and internal medicine physicians (L.I.H., G.Y.J., N.R.S.). Several participants were experienced in qualitative analysis (L.I.H., T.M., J.D.S.). We used a grounded theory approach, in which codes were generated inductively according to reading and rereading of the primary data.22 First the full study group individually read all the transcripts and met to discuss common themes and ideas generated from the first reading. A code list was then developed, and 3 study investigators (T.M., G.Y.J., L.I.H.) independently reviewed each reported incident to assign codes. At subsequent coding meetings, codes were generated, eliminated, or refined as needed with the constant comparative method.22 Disagreements were resolved by negotiated consensus. We continued this process until thematic saturation was achieved, ie, no new concepts were being generated. The full study group periodically reviewed the code structure. The final code structure included 35 unique codes, organized into 11 broad themes. We used Atlas.ti 5.0 (Berlin, Germany) to facilitate qualitative analysis.

We used descriptive statistics to characterize the data. Interrater reliability among the 3 coders, as assessed with the κ statistic, was excellent for type of information not communicated and consequences of failed transfers (all but 1 κ≥0.7).23 We compared categorical results among groups with the χ2 test. We used SAS 9.2 (SAS Institute, Inc., Cary, NC) to facilitate quantitative analysis.

Characteristics of Study Subjects 

We received a total of 139 of 264 responses (53%). These included responses from 39 of 60 ED house staff and physician assistants (65%), 21 of 37 hospitalists (57%), and 79 of 167 internal medicine house staff (47%).

Main Results 

Of the 139 respondents, 40 (29%) reported that a patient of theirs had experienced an adverse event or near miss after the ED-floor transfer (5 ED, 8 hospitalists, 27 internal medicine house staff). These 40 respondents described 36 specific errors, which were evenly divided among errors of diagnosis (N=13), treatment (N=14), and disposition (N=13). Examples of each are shown in Table 1. Some errors fell into more than 1 category. Physicians described 6 cases in which patients required an upgrade in care from the floor to ICU within 24 hours of admission.

Table 1. Representative examples of error types.

	Type of Error (Number of Incidents)	Reported Examples
	Diagnostic (13)	Acute multiarticular gout misdiagnosed as chronic leg pain
		Pulmonary embolus misdiagnosed as pneumonia
	Treatment (14)	Diabetic ketoacidosis not treated with insulin, fluids
		Sepsis not treated with antibiotics, fluids
	Disposition (13)	Patient with acute abdomen, perforation sent to medical floor; transferred to operating room
		Hypoxic, hypercarbic patient sent to medical floor; transferred to ICU

Despite the fact that we had worded the survey question to elicit reports of communication errors between ED and internal medicine physicians, analysis of the reported incidents made it clear that there were numerous vulnerabilities in the transfer process from ED to internal medicine, some of which did not involve communication at all. We identified vulnerabilities in communication, environment, information technology, patient flow, and assignment of responsibility. Each is described in more detail below and summarized in Table 2.

Table 2. Vulnerable components of the ED-floor transfer.

	Vulnerabilities	Potential Solutions
	Communication
	Communication omitted altogether	Automatic alerts to receiving physicians about new patients
	Key content omitted	Guidelines to encourage high-level content such as clinical assessment, certainty of provisional diagnosis, anticipatory guidance, pending studies
	Interaction limited	Adequate time for interaction; prioritize sign-out process; encourage synchronous communication
	Limited intragroup communication (ie, physician-nurse) and feedback (ie, IM to ED)	Formalized intragroup communication—physician-nurse rounds, ED-IM conferences, joint physician-nurse handoffs, mechanism for feedback
	In-group versus out-group bias—mistrust, stereotyping	Increased individual cross-functional expertise (IM rotations in ED; ED rotations on IM floors); encourage informal intergroup activities, collegiality; discourage culture of blame, disparagement
	Conflicting expectations about certainty, completeness, thoroughness of evaluation	Clarification of expectations
	Environment
	Crowding	Attention to physical design, workflow of ED and floors
		Improved hospital throughput
		Discharge time of day
		Bed turnaround time
		Elective operating room scheduling
		Transport
		Bed assignment
	High workload	Real-time work redistribution
		Staff to match expected demand
		Implement nontraditional inpatient admitting schedules (few admissions each day) instead of traditional inpatient call schedules (many admissions every few days) to reduce competing demands on admitting physicians
	Information technology
	Vital signs, ED notes, orders not electronic or easily accessible	Improved electronic access to key information
	Responsible physician not obvious to ancillary and specialty staff	Real-time update of responsible physician and means of contact Pager forwarding to facilitate role-based rather than individual-based responsibility
	Inpatient units unaware of ED crowding or delays	Bed management or throughput “dashboard” visible to inpatient and ED staff
	Patient flow
	Patients rebooked to new unit or patients going to dialysis before admission	Intermediate caregivers provide sign-out Patient returns to ED before going to new bed Sign-out written or recorded for durable access
	Direct admissions to medicine	Automatic alerts to receiving physicians about new patients
	Boarding patients in ED	Improved throughput
		Clarification of responsibility
		Focus on patients at greatest risk (requiring ongoing intervention)
		Delayed sign-out until bed available
	Assignment of responsibility
	Follow-up of pending data	Emphasis on role-based rather than individual-based responsibility
		Pager forwarding
	Responsibility for boarding patients	Mandated updates for boarding patients (within ED and ED-IM)

IM, Internal medicine.

Communication failure at some point of care was central to most, though not all, reported errors. The type of information not communicated included aspects of the medical history (N=7), physical examination (N=19), ancillary tests (N=9), and ED course (N=5). See Table 3 for examples of each. In 5 cases, no sign-out was conducted at all. Failure to communicate the most recent set of vital signs was the most prevalent specific content item; it was cited in 10 of 36 incidents (28%).

Table 3. Type of content involved in communication errors.

	Content Area (Number of Incidents)⁎	Example as Reported by Respondent
	History (7)	“Patient booked to the floor…who was in fulminant hepatic failure….No Tylenol level was done. It was assumed that because he is a known alcoholic this was his baseline liver function, which was erroneous. Also he did not have an alcohol level done, though he's required MICU care for delirium tremens in the past….The patient was transferred from the floor to the MICU the next day and has been there for approximately 4 days now.”
	Physical examination (19)	“60 Year old female with HIV and CD4 count indicative of AIDS. Patient was placed in the hall of the ED. Sign-out was given to the floor team and included the following criteria; fever, WBC, [anion] gap acidosis, mental status changes. When asked if the patient was safe enough to go to the floor, the admit BP 140/80 was given by the ED staff as a rationale that the patient was hemodynamically stable. Uncomfortable with the sign-out information, the floor team went to the ED and evaluated the patient. She was alone in a corner bed and her BP cuff had been cycling every 15 min with sequential systolic blood pressures of 140, 125, 110, 100, 94…. Within five minutes … the patient VF arrested. Luckily, the patient did eventually go to MICU and survive to discharge.”
	Ancillary tests (9)	“A young patient was admitted to me with acute hepatic failure who had an anion gap of 5 9 hours before I was called and a lactic acid of 10 one hour before sign out. These results were not known or followed up by the ED provider. The patient was tachypneic to the 30s upon arrival to the floor.”
	ED course (5)	“Patient had been receiving multiple liters of fluid of normal saline with no recheck of hypernatremia for which he was admitted. The sodium reported to the on call medicine team was [higher] than patient's actual sodium and emergency physician thought patient had received 1 L when he had received 3, going on 4 L of fluid with no recheck of sodium…which had dropped 8 points over 3 hours. Thankfully fluids were stopped on the floor and patient suffered no adverse effects.”

MICU, Medical intensive care unit; WBC, white blood cell; BP, blood pressure; VF, ventricular fibrillation.

⁎ The 5 incidents of completely omitted sign-out are not included. Ns sum to more than 31 because some incidents involved omission of more than 1 content area.

The quality of the final handoff between ED and internal medicine providers depended in part on other providers' previous conversations about the patient.24 In particular, respondents' comments suggested 3 main dyads of communication: ED provider-patient on entry to the ED (see for example, Table 3, “History”), ED provider-ED provider during the patient's ED stay (Table 3, “ED course”), and ED-provider-receiving physician on transfer out of the ED (Table 3, “Physical examination”). Misinformation or omission of data at any point in the sequence could propagate forward to the final, receiving internist. Conversely, clarification or correction during communication at any point could avert an error. Our analysis revealed a fourth dyad of communication failure: postevent feedback from receiving physician to emergency physician. Only 13.5% of ED respondents identified an adverse event occurring to a patient after ED-floor transfer, compared with 38.1% of hospitalists and 38.6% of internal medicine house staff (P=.02). Indeed, one ED provider commented, “[Adverse events for my patients because of transfer are] not applicable as far as I know,” which suggested that ED providers were rarely informed of problems experienced by patients once they had left the ED.

Interactive communication was also critical to the quality of the transfer. As one respondent observed, “Appropriate floor sign-out requires sufficient time for question/answer to verify appropriateness of disposition.” When present, discussion among participants could result in provision of additional information, clarification of misunderstood information, or reevaluation of diagnosis, management, or disposition. For example, in this case, an interactive discussion between emergency medicine and internal medicine residents made a diagnostic error apparent, allowing the treatment course to be appropriately altered: “Patient was signed out for ‘DKA' [diabetic ketoacidosis] even though patient's main problem was that he was in septic shock. This problem was addressed by medicine and ER residents before adverse outcome occurred.”

Similarly, in the example in Table 3, “Physical examination,” the information conveyed in the final handoff prompted the receiving team to evaluate the patient immediately, thus averting a potential fatality.

Group affiliation had a number of effects on communication quality.25 First, communication across group boundaries required extra effort and consequently was often omitted altogether. For example, the incident in which an emergency physician was unaware of events that had occurred in the ED (Table 3, “ED course”) appeared to be aggravated by a lack of communication with the ED nurses, who keep track of medication administration.

Second, in-group/out-group bias could lead to misunderstanding or mistrust between groups, impeding communication even when it did occur. Several internists, for example, mistrusted ED ability, judgment or professionalism, projecting their own perspective onto emergency physicians and privileging their own ability over that of others. After describing the incident in Table 3, “Physical examination,” one respondent wrote, “The above event highlights how imperative it be that ED staff recognize SIRS and severe sepsis criteria. If the floor team had not come down to evaluate that patient, I have no doubt she would have died in that hall bed in the corner before anyone recognized there was a problem.” This resident attributed a complex situation entirely to ED incompetence, ignoring the potential contribution of workload, information technology, physician-nurse communication, or “boarding.” Another respondent described a case of incomplete history-taking and then added, “I believe this scenario highlights that on a busy day, ED staff don't even speak with their patients, if they think they can get away with it.” This resident assumed a lack of professionalism, despite no information on actual emergency medicine actions.

Third, group membership conveyed particular expectations about diagnosis, management, and certainty that could obstruct mutual understanding. For example, one hospitalist reported, “A patient with a huge PE was signed out to me as a pneumonia.” The internist categorized this as a diagnostic error conveyed through sign-out that led to incorrect care on the inpatient side. Emergency physicians, however, typically view their role as stabilization and disposition, not definitive diagnosis and management. They would categorize this patient as appropriately admitted to medicine, with a tentative diagnosis. Likewise, conflicting expectations about management contributed to this incident reported by an internist: “I recently received a patient with a full code status on the floor who arrived on a regular medical floor with a non-rebreather mask on as she was actively seizing. The appropriate decision would have been to change this patient's code status if appropriate (which was ultimately done) or to admit her to the ICU.” Here the internist considered the absence of a code status discussion to be inappropriate; however, there are situations in which it is appropriate to defer such discussions, such as when a health care proxy or trusted loved one is not yet present in the ED.

High workload for sign-out participants, such as when the ED was crowded or when admitting physicians received numerous new patients, generated numerous competing demands on their time. This had several effects on patient care.

First, communication occasionally was omitted altogether. As one respondent described, “No sign-out was given on the patient. When nursing staff called me to evaluate, patient was tachypneic and tachycardic. In short…patient was transferred to CCU [cardiac care unit] with acute coronary syndrome. ED stated that this was an error secondary to being very busy with crowding in ED.”

Second, when participants were busy, sign-outs were reported to be rushed and less interactive, reducing opportunity to clarify information or thought processes.

Third, busy ED providers sometimes had not recently seen the patients they were signing out and were not aware of new developments or current vital signs, as occurred in this case: “Patient admitted to Generalist floor on busy ED time ‘red zone' [long waiting time for level 3 patients to be treated]; signed out as stable vitals but arrived cyanotic, PO2 50's with PCO2 in 80s. Rapid MICU transfer.” In this case the patient's condition appeared to have changed dramatically between the time the ED provider last treated the patient and the time he or she provided sign-out to the accepting team.

Fourth, when admitting physicians were themselves busy, they did not always treat patients immediately on transfer. As one internist reported, “Patients often have hemodynamic instabilities or lab abnormalities that are not mentioned in the very brief sign-out that's done, which leads to dangerous situations on busy call nights on things that may not be picked up by the call team until later.”

Finally, crowding not only increased workload but also caused some patients to be physically removed from standard care locations, increasing the possibility of being overlooked by physicians and nurses. The incident in Table 3, “Physical examination,” for example, concerns a patient who is in the “hall of the ED,” “alone in a corner bed.”

Thus, a crowded and chaotic environment coupled with high workload contributed to omitted, incomplete, or out-of-date sign-outs and led to delayed evaluations both within the ED and on the inpatient floor.

Information technology was implicated in several errors. Vital signs recorded in the ED were not visible electronically either to ED or IM providers. This contributed to numerous problems with out-of-date or inaccurate vital signs, the most commonly described content item leading to transfer-related problems and a major contributor to incorrect assessments of the patient's clinical condition.

Another recurrent problem was neglect of laboratory or test results that returned after the sign-out had taken place. As one respondent wrote, “[adverse events are] usually related to imaging studies completed just as patient in ED, but later sent to floor without calling or action related to imaging finding.”

Information technology contributed to this problem in 2 ways. First, inpatient providers were not always aware of pending studies because orders written in the ED were difficult to find in the electronic medical record. Second, laboratory and radiology personnel sometimes contacted the wrong provider with urgent results because responsibility for patients was in transition, as was their physical location. Yet there was no real-time, accessible means of identifying these changes through the information system. A series of messages was then typically required to get the results to the correct provider, with multiple opportunities for failed communication.

The nonlinear nature of patient flow through the ED threatened a transfer process that was designed for a linear progression of patients from triage to ED to floor. For example, some patients were triaged in the ED and then sent to dialysis. Often by the time dialysis was completed a bed had been found for the patient, who was then transferred directly to the floor. As a result, diagnosis and treatment could be omitted or delayed because the patient was physically isolated from both ED and inpatient providers. Furthermore, this created an opportunity for sign-out either to be omitted altogether or to be provided by emergency physicians who had not treated the patient in some time because intermediate providers, such as dialysis fellows, did not assume responsibility for sign-out. In this example, a patient in this situation experiences harm: “Ovarian torsion was missed because the patient went straight to dialysis without sign-out and then it was only later noted that the pelvic exam was markedly positive for cervical motion tenderness, and the patient had abdominal pain.”

A similar situation arose when patients were admitted to a team and then reassigned without notification to the emergency physician. In these cases, even if emergency physicians provided sign-out to the initial team, the final admitting physician would not receive any sign-out. Again the intermediary—the initial receiving team—did not assume responsibility for sign-out to the reassigned team. Thus, the transfer process failed, even though information may have been perfectly communicated between the ED and the initial team. An example of this type of situation was reported by an internal medicine resident: “A patient with chemotherapy-induced diarrhea and hypokalemia with EKG changes was signed out to and accepted by the oncology service. The only telemetry bed available for the patient was on an internal medicine floor to which my team admitted patients. The patient was subsequently admitted to the medical floor from the ED and arrived on the floor without the internal medicine team knowing who the patient was, what her diagnosis was, the severity of her hypokalemia, or need for telemetry monitoring.”

A third situation of nonlinear patient flow involved patients who had been admitted but whose bed was not ready. In these situations, patients were typically signed out to the admitting team, yet continued to “board” in the ED until the bed was available. Consequently, communication was uncoupled from the physical transfer. Policy dictated that emergency physicians maintain primary responsibility for boarding patients. In practice, however, emergency medicine providers tended to move on to take care of new patients once their patient had been signed out. In addition, they did not typically sign out again to the admitting team, even if the patient remained in the ED for several more hours, and at times provided only cursory sign-outs to their oncoming colleagues in the ED if a shift change intervened. Boarding patients were thus in limbo, placing them at risk of not receiving care from either emergency medicine providers or inpatient providers. This danger was exacerbated by the fact that boarders tended to accumulate in the ED precisely when the ED was busiest and emergency physicians had the most competing demands on their time. As one respondent described, this was particularly problematic for patients who required continuing intervention and treatment, such as “diabetics not receiving adequate fluid or insulin, and asthmatics not continuing to get treatment while waiting for a bed.”

Finally, emergency physicians had admitting protocols in place with internal medicine such that official action by internal medicine physicians was not necessary for the ED to have a patient admitted. This improved ED patient flow and ensured that every patient had an inpatient “home.” However, it also meant there was no forcing function to ensure that internal medicine providers knew about every new patient. Without a checkpoint, patients sometimes fell through the cracks and were “admitted to medicine without sign-out (just showed up on the floor).”

Ambiguous assignment of responsibility also contributed to transfer errors. For example, studies were often ordered by the emergency physician after sign-out at the request of the admitting physician, generating confusion about who was responsible for following up results: the ordering physician or the one who had requested it be ordered. In addition, as previously discussed, ambiguous responsibility for sign-out and patient care was problematic for boarding patients and for patients who did not travel directly from the ED to their final inpatient bed.