Transfers of patient care from one physician to another, a process known as "handoffs," are pervasive in health care (1). These handoffs are often accompanied by communication and other failures (2, 3), which can lead to medical errors and harm to patients (4).
Research to date focuses on transitions necessitated by transfer to a different setting of care (e.g., transfer from a general ward to an ICU) (3) or by the end of a shift (2). A care transition that has received only recent attention in the patient safety literature, but may well entail significant risks, is the change in personnel that occurs in the outpatient clinics of academic health centers on July 1st when residents either advance to a higher level of training or graduate (5,6).
While year-end handoffs resemble other types of handoffs in many regards, there are important distinguishing features. First, the year-end transfer terminates a longer-term treatment relationship which in turn can be experienced by patients as a significant loss leading to increased symptoms (7). This heightened acuity makes monitoring of higher-risk patients even more critical during the transition. Second, residents stationed in continuity clinics have often accumulated relatively large panels of patients. As a result, the year-end transfer requires unique organizational processes capable of simultaneously transitioning a high volume of patients. Without such procedures in place, there is a much higher probability for patients or important clinical information "falling through the cracks" compared with other types of transfers. Third, patients are transferred from trainees with substantial experience to those with much less (outpatient) experience and skill. This experience gradient can be significant and warrant slower caseload growth, enhanced supervision, and specialized didactics (5).
This article describes a patient suicide in the context of a year-end transfer and the findings of a subsequent root cause analysis. We identify the sources of error in the year-end transfer process that contributed to the suicide and then propose strategies to enhance patient safety and resident training.
Approximately 400 patients are transferred in the Medication Management Clinics on July 1 from graduating or fourth-year residents to incoming third-year residents. A paper chart system coexists with an electronic billing and scheduling system. The incoming residents have no prior experience in this institution's outpatient department and have spent the majority of their first 2 years of training in inpatient environments.
Ms. B, diagnosed with recurrent major depressive disorder, experienced increased depressive symptoms, suicidality, and alcohol abuse in the spring. She refused higher levels of care and never met grounds for an involuntary hospitalization. The outgoing resident last saw her in clinic on June 5 during which the patient showed some signs of improvement. The plan was for Ms. B to continue weekly psychotherapy with her therapist in the community and return to the Medication Management Clinic in July when she would meet her new psychiatrist. The resident did not discuss this plan with the community therapist.
Subsequent to this visit, Ms. B met with her therapist as planned. Ms. B reported increased depression to the therapist as the anniversary of her infant's death approached. The therapist had also made arrangements to transfer the patient to another therapist starting in July. No communication occurred between the resident and therapist during this transition period. As a result, the resident was not aware of the anniversary, the increased symptoms, or the simultaneous transfers. A new third-year resident assumed care of the patient on July 1. No Medication Management Clinic appointment had been made and no protocol existed whereby acute or higher-risk patients were identified and either contacted or scheduled on a priority basis. At some point in mid-July, Ms. B committed suicide.
Ms. B's death represented a sentinel event. A root cause analysis followed. Participants included clinic leadership, the patient safety officer, two clinicians from other services, and the administrative supervisor for the Medication Management Clinic. The team developed a list of human and other factors that contributed to the adverse event and identified the related processes and systems. For each contributing factor, root causes were identified through a series of "why has the factor occurred" questions. Solutions were then designed to prevent future recurrences. Institutional review board approval was obtained for purposes of publication.
The root cause analysis identified a number of factors contributing to the patient's suicide, including the patient's decision to not engage in a higher level of care, comorbid alcohol abuse, two concurrent transfers, the anniversary of the death of the patient's infant, inadequate communication between the two principal clinicians in the context of recent symptoms and impending transitions in care, no standardized handoff of clinical information from the outgoing to incoming resident, and no mechanism by which incoming residents, the supervisors and the clinic prioritize scheduling of more acute or higher risk patients. After this initial assessment, the root cause analysis focused the problem definition on the "event of a suicide in the context of multiple transfers of care." Because many of the errors related to an inadequate year-end transfer process, subsequent analysis focused exclusively on the transfer process. Three principal sources of error in the year-end transfer process were identified: unbalanced caseloads, communication failures, and inadequate transfer of professional responsibility. Each of these problems is discussed below.
For reasons of administrative simplicity, new caseloads largely preserved the prior caseloads (i.e., incoming resident A received all of outgoing resident B's patients). Trainees complained each year that the caseloads were uneven with significant variability in terms of acuity and complexity. To further assess this complaint, we compared resident productivity during the first 5 months of the academic year and found that the existing transfer process yielded a marked variability in the number of patients seen by different residents, with more than a two-fold difference between the lowest and highest. The clinic had a small number of patients whose clinical needs were much greater than what the clinic could offer. We asked each resident to estimate the number of nonclinic minutes per month they spent on each patient. This revealed an uneven distribution of "high need patients." Two residents had zero such patients while one resident had 16 patients requiring more than 60 nonclinic minutes per month. This variability was at least in part related to the method used to construct initial caseloads and likely contributed to some trainees having felt overwhelmed and unable to adequately manage the clinical tasks (6).
The analysis of how administrative staff processed the paper transfer requests revealed that multiple staff members were entering data in a nonstandardized manner into redundant databases without any data verification processes. These two databases often conflicted, and administrative staff differed in terms of which database they deemed as the standard. Consequently, patients were often assigned to more than one clinician or none.
In addition, the primary and often sole source of clinical information for the incoming resident was the transfer request document which was written 6 to 8 weeks prior to transfer. No verbal or face-to-face sign-out existed. Acute patients were not flagged. Acute patients or patients who did not manage transitions well were not necessarily seen in the first month of the new residents' clinics. Supervision efforts and didactics did not specifically prepare outgoing or incoming residents for the transfer. And, especially important in this case, communication with other members of the treatment team, including psychotherapists, did not occur during the transitional period, even for higher-risk patients.
+
Culture and Professional Responsibility
Finally, the analysis revealed that the residents' schedules were slowly ramped up from two patients per clinic to five over the first 3 months. This policy had been adopted several years earlier in order to help residents cope with the stresses associated with the significant increase in clinical responsibility and autonomy. Discussions with faculty and residents surfaced the concern that this "supportive measure" sent the unintentional signal that the "clinic" was taking care of those patients not seen, whereas, in fact, most of the patients in a given resident's panel are only seen by that resident. Moreover, the new third-year residents were primed to misinterpret this signal since their only prior outpatient experience took place in team-based clinics at a VA where the "clinic" held the patients.
Figure 1 portrays the temporal relationship of the errors by way of the "Swiss cheese model" of system errors originally introduced by James Reason (8). This model posits that the holes in the cheese represent errors that produce a bad outcome only when they line up (i.e., occur in close proximity to each other).
Multiple factors contributed to the death of this patient in the context of the year-end transfer, including factors not related to the transfer process such as the disease process itself (e.g., comorbid alcohol abuse and associated impulsivity), the patient's own choices about level of care, and clinician error (e.g., inadequate interdisciplinary communication, not aware of an important anniversary, insufficient follow-up plan). Even if a perfect year-end transfer process had existed, the suicide may not have been prevented. Nonetheless, the results of this analysis highlight the risks associated with year-end transfers that occur as a result of residents who graduate or advance.
Many of the findings from this analysis parallel what has been written about other types of handoffs. Our analysis identified a number of contributing communication failures, including nonstandard data entry leading to inaccurate clinical information and the absence of structured written and face-to-face sign-outs and reliable methods to identify and prioritize acute patients. These failures have been repeatedly observed in studies of the end-of-shift handoff (1—3, 9—11).
Similarly, several factors contributed to the incoming residents not fully and effectively assuming responsibility for their new panel of patients, including a need for training for handoff-related clinical tasks and the lack of perceived clarity around who is responsible for patients not yet seen. These problems have also been noted in studies of other types of handoffs, including the problem of agency in which physicians do not act in the best interest of their patients—a problem potentially compounded when physicians are responsible for patients they do not (yet) know (10).
+
Recommendations for Designing Safer Year-End Transfers
Given the significant risks associated with other types of handoffs, considerable attention has been focused on interventions to improve patient safety and reduce errors during transfers, many of which have been adapted from industries such as nuclear power and space aviation in which transition errors have high consequences (11). We have adapted findings from our analysis above and the literature on end-of-shift handoffs to propose features of a year-end transfer process that is safer for patients and a better educational experience for residents. Table 1 summarizes these proposed features.
No prior studies have addressed this type of handoff. Future research should examine practices across disciplines and institutions, medical errors and harm to patients that most commonly occur, and interventions that improve patient safety and resident education. Possible outcome measures include the impact on rates of clinical decompensation during the transfer period, the time to first visit for acute patients, and the quality of the handoff communication as perceived by the clinicians.
At the time of submission, the authors reported no competing interests.