Horizontal care, in which clinicians assume roles outside of their usual responsibilities, is an important health care systems response to emergency situations. Allocating residents and fellows into skill-concordant clinical roles, however, is challenging. The most efficient method to accomplish graduate medical education (GME) assessment and deployment for horizontal care is not known.
We designed a categorization schema that can efficiently facilitate clinical and educational horizontal care delivery for trainees within a given institution.
In September 2019, as part of a general emergency response preparation, a 4-tiered system of trainee categorization was developed at one academic medical center. All residents and fellows were mapped to this system. This single institution model was disseminated to other institutions in 2020 as the COVID-19 pandemic began to affect hospitals nationally. In March 2020, a multi-institution collaborative launched the Trainee Pandemic Role Allocation Tool (TPRAT), which allows institutions to map institutional programs to COVID-19 roles within minutes. This was disseminated to other GME programs for use and refinement.
The emergency response preparation plan was disseminated and selectively implemented with a positive response from the emergency preparedness team, program directors, and trainees. The TPRAT website was visited more than 100 times in the 2 weeks after its launch. Institutions suggested rapid refinements via webinars and e-mails, and we developed an online user's manual.
This tool to assess and deploy trainees horizontally during emergency situations appears feasible and scalable to other GME institutions.
Horizontal care, in which clinicians assume roles outside of their usual responsibilities, is an important health care systems response to emergency situations. An efficient method to mobilize trainees during emergency situations has not been studied.
A categorization schema that can efficiently facilitate clinical and educational horizontal care delivery for trainees within a given institution.
The categorization scheme broadly classifies residents and fellows into only 4 categories and may not reflect the exact needs of a specific disaster response.
A horizontal deployment classification scheme for residents and fellows was disseminated nationally and further refined during the COVID-19 pandemic. It was feasible, acceptable, and may be scalable to institutions of varying sizes.
Over the past 2 decades, disaster events in the United States have necessitated rapid coordination and mobilization of our hospital workforce. Much of the planning for these disasters has centered around mass casualty incidents.1–4 However, the 2015 Ebola virus outbreak reinforced the importance of incorporating pandemic events within health care disaster planning,5–7 and the 2020 COVID-19 pandemic has again proven the need for effective systems-level preparations. Without established plans and logistical support, hospitals may struggle to identify personnel to fill vital roles in emergencies.
Residents and fellows have added substantial value to previous emergency response situations for several reasons.8,9 These physicians represent a sizeable, skilled workforce that is easily accessible, rapidly mobilized, and often geographically clustered. They are also adept at navigating electronic health record systems, allowing them to transfer to other roles and locations with minimal additional training. Working on different services and in different departments throughout a hospital system affords them an intimate working knowledge of hospital policies, geography, and other personnel. This is a distinct benefit when compared to locum or other transient clinicians who would require orientation and training. Lastly, and perhaps most importantly, trainees are an intelligent and high-functioning group, are well-educated, and can be retrained quickly. Some are independently licensed, and most fellows are board-eligible or board-certified, allowing for their emergent credentialing if necessary.
An efficient method to mobilize trainees during emergency situations has not been studied. Therefore, in September 2019, we sought to create an emergency response system to rapidly communicate with and horizontally deploy them for a variety of disasters. We describe our process and subsequent collaboration with the University of Michigan that allowed for free implementation of this system to support national COVID-19 pandemic response efforts.
Setting and Participants
Vanderbilt University Medical Center (VUMC) is a teaching hospital in Nashville, Tennessee, with 98 Accreditation Council for Graduate Medical Education (ACGME)–accredited programs and 1053 residents or fellows appointed through the graduate medical education (GME) office. All trainees are graduates of allopathic or osteopathic medical programs or their equivalent, and 22% have Tennessee medical licenses. Approximately 25% are in fellowships and are board-certified or board-eligible in their core specialties such as surgery, pediatrics, anesthesiology, or medicine.
In September 2019, the VUMC GME office brought together key stakeholders with expertise in disaster management and emergency preparedness to discuss institutional needs and determine relevant roles that trainees might fill in a health care emergency. These stakeholders determined the need for an inventory of trainee patient care skill sets according to specialty and postgraduate year (PGY) level.
A VUMC workgroup was subsequently formed and set out to develop a categorization system that would allow the GME office, GME administrators, and the hospital command center to have a rapid assessment of trainee workforce capabilities. The workgroup discussed different models of care delivery and referred to them as vertical and horizontal. Vertical care delivery is intradepartmental and can be coordinated quickly and easily. Horizontal care delivery, in contrast, refers to a comprehensive, institution-wide approach to mobilizing the existing workforce to meet immediate needs. Horizontal care delivery must consider clinician skills across departments and requires not only immediate action but also a response that would be efficient, robust, and sustainable.
Development of Classification Schema
Residents and fellows were broadly categorized by core specialty areas (medicine, surgery, pediatrics, anesthesia) and then, within each specialty, by clinical skill level corresponding to expected ACGME Milestone advancement by PGY. We set a 4-tiered ranking system: Independent, Advanced, Basic, and Foundation. This system was designed to disassociate skills from individual trainees and instead advance trainees yearly as they advance in their programs.
The workgroup also determined the need for additional medical care categories beyond the 4 broad specialty areas, given patient and health system needs during an emergency. A Family Support category was created as previous incidents have identified the need to provide support for both patient families and health care workers.10,11 Certain medical specialties were also identified as having unique skill sets such as radiograph interpretation, laboratory analysis, and anesthesia administration and were thus grouped into Specialized Services. Finally, an Intensive Care Unit category was created to account for critical care skills for both adult and pediatric patients.
Trainees were categorized based on their ACGME Milestone level and assumed skill set on July 1 of a given year. Although they are certainly more advanced in March of any given academic year than July, the system was designed to be operational at any time point and for any emergency context. Thus, those trainees not in Independent, Advanced, Basic, or Specialized skill categories on July 1 (eg, first-year residents who have just entered a program and must be under direct supervision) can augment nonclinical tasks essential to patient care (such as transport, central supply, meal service, etc) and were therefore assigned to the Foundation category. Table 1 depicts how these 4 skill categories were then linked to trainee disaster roles.
Institutional GME Assessment
Next, an institutional GME assessment was performed. The VUMC GME office e-mailed all 98 program directors in December 2019, with explanations of assessment goals, the categorization schema, and a survey requesting that they confirm the classification assigned to their trainees. As aforementioned, program directors were instructed to assume the date to be July 1 of any year, considering expected Milestone progression. With 100% participation and 100% of trainees categorized, we obtained a cross-sectional picture of our institution's GME workforce capabilities.
Rapid Communication System Implementation
Previous disasters such as the Boston Marathon bombing have led to shutdowns of cellular communication systems in affected areas.2,12,13 As a result, a reliable and rapid means of communication is critical in disaster response. VUMC utilizes a mass notification platform (Everbridge Critical Event Management, Pasadena, CA) for time-sensitive situations such as mass casualty incidents and inclement weather. Through this platform, VUMC can send information in several formats (text, phone call, or e-mail) to all employees in seconds. Message groups can also be created to transmit specific updates to a subgroup of employees. In January 2020, trainees were allocated into message groups based on their skill categorization and year of training. This allows the GME office to centrally and rapidly mobilize trainees with specific skills or experience levels depending on the needs of the hospital.
Multi-Institutional Implementation During COVID-19 Pandemic
Recognizing that all institutions were facing potential surges in patient admissions due to the COVID-19 pandemic, the VUMC designated institutional official (DIO) shared the VUMC model via a listserv for the Association of Program Directors in Surgery in March 2020. The DIO was subsequently approached by colleagues at the Center for Surgical Training and Research (University of Michigan) and the Procedural Learning and Safety Collaborative who expressed interest in adapting the tool to be specific to the COVID-19 pandemic. The COVID Staffing Project,14 a multi-institutional collaborative, was immediately formed to implement this system locally at VUMC as well as support GME programs broadly.
To translate the VUMC classification system into a scalable solution for other institutions, an interdisciplinary team of physicians, data analysts, engineers, software developers, and social scientists adapted the VUMC schema and trainee disaster roles into a matrix of COVID-specific Response Roles (table 2). Attending physician as well as trainee roles were categorized because some hospitals may choose to put emergently credentialed fellows into attending physician roles. This team then created the Trainee Pandemic Role Allocation Tool (TPRAT)14 and made it available for free download on March 31, 2020. The TPRAT is a layered spreadsheet built in Microsoft Excel (version 16.35, Redmond, WA) that automates the organization of trainees into COVID-19 clinical roles based on their specialty and year of training. An overview of how to use the TPRAT is provided in table 3, figure 1, and figure 2. We estimate that approximately 300 hours of collaborative work went into the development of this tool.
Outcomes and Analysis
The emergency response system was intended for use during 3 disaster scenarios: (1) mass casualty; (2) environmental or structural disaster; and (3) pandemic. In March 2020, VUMC experienced 2 of these scenarios: Nashville, Tennessee, was hit by a tornado on the evening of March 3, and the COVID-19 pandemic unfurled in the subsequent weeks. Responding to both situations necessitated selective horizontal deployment of clinicians across VUMC and allowed for initial assessment of the feasibility, acceptability, and usefulness of the system. The tool was also shared nationally in the United States.
This work was considered a quality improvement initiative and as such was not reviewed by the Vanderbilt University Medical Center Institutional Review Board.
The horizonal deployment model was referenced to determine which of our trainees should remain in reserve during the weeks prior to a potential COVID-19 surge. It was subsequently used to create proactive call schedules for procedure teams as well as outpatient and intensive care unit surge teams to ensure supervision, maximize educational value, and oversee transfer of skills. As the framework had already been established, it was very easy to implement. The DIO added a system in which program directors could indicate presence or absence of trainees in a shared folder for real-time assessment of available trainees. As all program directors had been directly involved in its creation and understood the basics even before the COVID-19 pandemic, the implementation was very well received. Additionally, because the framework had been established, efforts could be focused on communication regarding the well-being of trainees.
When VUMC shared the model via an ad hoc e-mail listserv with 32 DIOs during the week of March 23, 2020, several requests were made to review the classification scheme. Approximately 7 DIOs from national institutions participated in an initial informal webinar. Several subsequent individual videoconferences and conversations were requested, and at least 2 other institutions adapted the original model.
Two weeks after the online launch of the TPRAT (March 31, 2020), the site was shared via a listserv for the Group on Resident Affairs of the Association of American Medical Colleges. The COVID Staffing Project website has received more than 100 visitors as of April 14, 2020. Administrators have asked questions about the tool via informational webinars or e-mail, and we have incorporated several changes to the TPRAT based on their feedback. As one example, the entire list of ACGME subspecialty residencies was added to the input worksheet while preserving the capability to enter non-ACGME trainee data, which improves the TPRAT's customizability for all institutions. Other improvements were also made to the tool and its accompanying User Guide.16 Finally, we refined the TPRAT to work synergistically with other tools, including those of the COVID Staffing Project, to maximize the integration of trainees into hospitals' broader COVID-19 response efforts.
This model for horizontal deployment of trainees during a variety of emergency situations, which categorized trainees into Independent, Advanced, Basic, and Foundation categories, was acceptable during the COVID-19 pandemic at 1 institution. With multi-institutional input, the instrument was refined and developed into a COVID-19 specific tool (the TPRAT) for categorizing trainee specialties, with broad interest nationally.
Several other issues must be considered in horizontal care delivery, particularly by trainees. Safety and educational value must be upheld alongside high-value patient care. Trainees should always be in the most relevant educational and clinical roles possible, with the goal of maximizing learning and contribution during a disaster if they are horizontally deployed. GME programs should look to the ACGME for guidance, an example being the “Stage 3: Pandemic Emergency Status Guidance” that the organization published during the COVID-19 pandemic.17 In this guide, nonnegotiable considerations were outlined, including the need to continue to provide adequate resources and training, adequate supervision, and adherence to work hours.
Process sustainability is an important long-term consideration. Since skill categorization was based on expected milestones associated with each PGY, the categorization of each individual resident can be advanced as the resident matriculates through training, creating a durable process. Additionally, the TPRAT output spreadsheets can be easily modified, allowing program directors, administrators, or command center leaders to track trainee availability and adjust deployment accordingly as emergent situations such as the COVID-19 pandemic evolve.
There are some limitations to this system. The categorization scheme broadly classifies residents and fellows into only 4 categories and may not reflect the exact needs of a specific disaster response. For example, the relative lack of emphasis on surgical services during a COVID-19 pandemic response plan would not align with needs following a mass shooting, and thus each disaster may require customization. Also, this process categorizes residents on their level of training, rather than their individual performance, which may overestimate or underestimate their skills. We have shown that the categorization process has wide appeal, but we have not examined actual patient or educational outcomes, in comparison to other strategies to horizontally deploy trainees. As the original categorization scheme was developed for a large institution, the generalizability to GME institutions with low numbers of trainees or residency programs is unknown.
Since implementation of the classification scheme at VUMC, proactive schedules for intensive care unit surges and procedure teams have been created 2 to 4 weeks in advance, with daily discussions regarding program pressures and trainee availability. How residents are deployed horizontally during the pandemic, versus schedules generated from the classification scheme, should be followed during the pandemic. A model for horizontal redeployment has been created for faculty, with data from a skills survey, and its implementation should be followed as well.
This horizontal deployment classification scheme for residents and fellows, developed initially for one large institution, was disseminated nationally and further refined during the COVID-19 pandemic. The tool appears feasible and widely acceptable and may be scalable to institutions of varying sizes.
Denotes co-first authors.
Bradley Dennis, MD*, is Interim Chief, Division of Trauma and Surgical Critical Care, and Associate Professor, Department of Surgery, Vanderbilt University Medical Center (VUMC); Alexandra Highet, BA*, is a Fourth-Year Medical Student, University of Michigan Medical School; Daniel Kendrick, MD, MA, is Clinical Instructor of Surgery, Department of Surgery, Michigan Medicine; Laura Mazer, MD, MHPE, is Assistant Professor of Surgery, Michigan Medicine; Sean Loiselle, BS, is Full Stack Developer, Procedural Learning and Safety Collaborative; Hoda Bandeh-Ahmadi, PhD, is Director of Social Research, Center for Surgical Research and Training, Department of Surgery, Michigan Medicine; Tanvi Gupta, ME, is Big Data Analyst, Center for Surgical Research and Training, Department of Surgery, Michigan Medicine; Kenneth Abbott, MS, is a Fourth-Year Medical Student, University of Michigan Medical School; Jarrett Lea, BS, is Senior Application Developer, Office of Graduate Medical Education, VUMC; Thu Dang, BS, BA, is Associate Application Developer, Office of Graduate Medical Education, VUMC; Mischon Ramey, MS, is Senior Application Developer, Office of Graduate Medical Education, VUMC; Brian George, MD, MA, is Assistant Professor of Surgery, Department of Surgery and Director, Center for Surgical Training and Research, Michigan Medicine; and Kyla Terhune, MD, MBA, is Associate Professor of Surgery and Anesthesia, Vice President for Educational Affairs, Associate Dean for Graduate Medical Education, and Accreditation Council for Graduate Medical Education/National Resident Match Program Designated Institutional Official, Department of Surgery, VUMC.
Funding: The authors report no external funding source for this study.
Conflict of interest: The authors declare they have no competing interests.
The authors would like to thank John Morris, MD, Associate Chief of Staff at VUMC, and Donald Brady, MD, Senior Vice President for Educational Affairs at VUMC, for their support, the program directors at VUMC for their verification of training roles and contributions to the model, and all frontline and supportive health care clinicians worldwide in the COVID-19 pandemic.