Practical Radiation Oncology (2011) 1, 2–14

www.practicalradonc.org

Special Article

The challenge of maximizing safety in radiation oncology Lawrence B. Marks MD a,⁎, Marianne Jackson MD, MPH a , Liyi Xie MD a,b , Sha X. Chang PhD a , Katharin Deschesne Burkhardt MS, DABR a , Lukasz Mazur PhD c , Ellen L. Jones MD, PhD a , Patricia Saponaro MS a , Dana LaChapelle RTT a , Dee C. Baynes RN, BSN a , Robert D. Adams EdD, CMD a a

Department of Radiation Oncology, University of North Carolina, Chapel Hill, North Carolina Department of Radiation Oncology, Fudan University Shanghai Cancer Center and Department of Oncology, Shanghai Medical College, Fudan University, Shanghai, China c Industrial Extension Service, North Carolina State University, Raleigh, North Carolina b

Received 28 September 2010; accepted 4 October 2010

Abstract There is a growing interest in the evolving nature of safety challenges in radiation oncology. Understandably, there has been a great deal of focus on the mechanical and computer aspects of new high-technology treatments (eg, intensity-modulated radiation therapy). However, safety concerns are not limited to dose calculations and data transfer associated with advanced technologies. They also stem from fundamental changes in our workflow (eg, multiple hand-offs), the relative loss of some traditional “end of the line” quality assurance tools (port films and light fields), condensed fractionation schedules, and an under-appreciation for the physical limitations of new techniques. Furthermore, changes in our workspace and tools (eg, electronic records, planning systems), and workloads (eg, billing, insurance, regulations) may have unforeseen effects on safety. Safety initiatives need to acknowledge the multiple factors affecting risk. Our current challenges will not be adequately addressed simply by defining new policies and procedures. Rather, we need to understand the frequency and causes of errors better, particularly those that are most likely to cause harm. Then we can incorporate principles into our workspace that minimize these risks (eg, automation, standardization, checklists, redundancy, and consideration of “human factors” in the design of products and workspaces). Opportunities to enhance safety involve providing support through diligent examinations of staffing, schedules, communications, teamwork, and work environments. We need to develop a culture of safety in which all team members are alerted to the possibility of harm, and they all work together to maximize safety. The goal is not to eliminate every error. Rather, we should focus our attention on conditions (eg, rushing) that can cause real patient harm, and/or those conditions that reflect systemic problems that might lead to errors more likely to cause harm. Ongoing changes in clinical practice mandate continued vigilance to minimize the risks of error, combined with new, nontraditional approaches to create a safer patient environment. Published by Elsevier Inc. on behalf of American Society for Radiation Oncology.

Conflicts of interest: Dr Marks has been an unpaid consultant for Varian, Siemens, and Impac. The Department of Radiation Oncology at University of North Carolina receives research support from Siemens. ⁎Corresponding author. Department of Radiation Oncology, Box 7512, University of North Carolina, Chapel Hill, NC 27514. E-mail address: [email protected] (L.B. Marks). 1879-8500/$ – see front matter. Published by Elsevier Inc. on behalf of American Society for Radiation Oncology. doi:10.1016/j.prro.2010.10.001

Practical Radiation Oncology: January-March 2011

Introduction Radiation therapy (RT) is generally safe, largely because of the decades-long recognition of its risks. This is a tribute to the founding members of our field who recognized the risks of RT and who instilled within the very fabric of our field the need for careful oversight and clinical observation. Furthermore, physicists, engineers, and other technical and quantitative-minded individuals, integral to our practice, bring an objective and systematic approach to quality assurance. Nevertheless, there are safety challenges within our field that are related, at least in part, to the increased complexity of advanced techniques. Radiation therapy safety was the topic of a series of recent articles in The New York Times1-4 and multiple professional journal articles,5-14 including a Red Journal Special Issue in 2008.15 In 2010, radiation therapy safety was the focus of Congressional hearings, a public meeting sponsored by the Food and Drug Administration, a “Call to Action" meeting (filled to capacity) jointly sponsored by the American Society for Radiation Oncology (ASTRO) and the American Association of Physicists in Medicine (AAPM), and ASTRO announced plans to address safety concerns (Appendix).16 Thus, there is growing interest in the evolving nature of safety challenges in radiation oncology. Herein, we will: 1) review data regarding the reported frequency of radiation therapy “errors”; 2) present a broad framework to explain the recent focus on safety matters; and 3) describe several concepts and initiatives to enhance safety, some that have been implemented at University of North Carolina (UNC) and elsewhere. The broad thesis of this summary is that much of the recent concern regarding safety understandably has been focused on the mechanical and computer aspects of new, high-technology treatments (eg, intensity-modulated radiation therapy [IMRT]). This is warranted as the implementation and use of new and complex techniques certainly can increase risks. Some new technologies also alter fundamental worker responsibilities, may instill an unwarranted perception of infallibility, and challenge some longstanding approaches to quality assurance (QA). Furthermore, there have been other changes in our workspace and the medical system in general that influence safety. Our current challenges will not be adequately addressed by new procedures. Rather, our field needs to better understand the frequency and causes of errors, particularly those that are most likely to cause harm. We also need to incorporate basic human factors principles that minimize risks into the design of our workspace and services. These basic principles will include things such as automation, standardization, checklists, workflow improvement, and redundancy for high-

Safety in radiation oncology

3

risk procedures. We need to develop a culture of safety in which all of the team members are working together to maximize safety, and in which safety initiatives acknowledge the “hierarchy of effectiveness.”

I. How common are errors in radiation treatment delivery? Estimating error rates is challenging. The definition of an “error” is ambiguous and underreporting is the norm. Even if a consistent definition is applied, there have been only limited attempts to estimate the rates of particular types of errors. The successful delivery of RT requires the concerted efforts of multiple individuals, and errors can occur anywhere in the process. James Reason developed the “Swiss Cheese Model” to describe the mechanisms by which human error occurs in complex organizations. He suggested that “holes” in the layers of controls have to align to allow errors to penetrate the barriers that are normally in place to prevent them (Fig 1). To identify the source of error, one must investigate each layer, including the policies and practices that might create vulnerability. In general, “upstream errors” (eg, in physician evaluation or communication to technical staff) are difficult to identify, often subjective, frequently corrected “downstream,” and are not widely reported. Errors in dose calculation and treatment delivery are by their nature more readily identified and objective. Errors in treatment delivery are essentially always manifested at the treatment machine, even if its “cause” was upstream (eg, ambiguous physician directive), or, as is often the case, the result of a series of errors throughout the process (Fig 1). Thus, most of the data regarding error incidence focuses on events occurring at the end of the process, with some attribution bias. Diligent root cause analyses are needed to identify deviations and frailties upstream.17,18 With these caveats stated, several data sources can be used to infer information on error rates (Table 1). 1. New York State has maintained a central registry of reportable radiation “events” between 2001 and 2009. Their database contains 230 “events” derived from an unknown number of patients and treatments. Using estimates based on radiation use and cancer incidence rates in the state of New York, one can estimate that the rate of reportable events is 230 of 373,000. Of these, 37 of 230 required medical intervention for a “serious event” rate of ≈1.0 of 10,000 treated patients, according to the New York State Department of Health estimates (J. Krishnamoorthy, personal communication, August 6, 2010). “Therapist error” was implicated in 84% of the events.19

4

L.B. Marks et al

Practical Radiation Oncology: January-March 2011

Figure 1 Simplified workflow diagram illustrating the approximate steps between consultation and treatment with external beam radiation therapy. Errors in treatment delivery can result from an error anywhere along the path. At any step in the process, errors occurring “up-stream” can be detected and corrected before impacting the patient (upper dotted line with error detected during plan approval). Because there are often many opportunities for errors to be detected, some errors become clinically manifest only after a series of errors throughout the process (solid line). Errors originating downstream have fewer “opportunities for correction” (angled solid line). Furthermore, because errors in treatment delivery are always manifest at the treatment machine (even if their genesis [at least in part] is more upstream), therapists are frequently implicated in delivery errors. Some quality assurance (QA) steps noted in the upper section have become somewhat less useful in the intensity-modulated radiation therapy (IMRT) era. CT, computed tomography; DRR, digitally reconstructed radiograph; DVHs, dosevolume histograms; Tx, treatment.

was “requested port film shifts” included as events. In concert, these studies suggest that approximately 1% of patients have at least a minor deviation during their course of treatment with the majority causing no lasting harm. Nevertheless, their occurrence might reflect underlying problems. As expected, this 1% rate is far higher than that estimated from the New York or United Kingdom series, because the threshold for reporting to an external agency is higher than that used in an institutional review.

2. A review from the United Kingdom noted 181 incidents during 6.25 years, corresponding to an estimated rate of 40 incidents (3 clinically significant) of 100,000 per courses of therapy.20 3. Several institutional series note deviation and error rates per patient, field, or treatment course (Table 1). Obviously, as one alters the denominator, the reported rate varies. The definition of a deviation was also inconsistent. Typically, all treatment deviations were included, whether or not they were clinically meaningful. In none of these studies Table 1

Estimated deviation rates from selected population and single institutional studies Deviation rates (%) Per treatment

Multiple center series UK20, 2007 Pennsylvania State50, 2009 NY State19, b, 2009 Single institution series Fraass6, 1998 Macklis9, 1998 Barthelemt-Brichant13, 1999 Patton11, 2003 Huang8, 2005 Yeung12, 2005 French7, 2006 Marks10, 2007 a

Per course

Per field

0.04 (0.003) 0.0025 (0.0006 a) 0.06 (0.01 a) 0.44

0.17

1.20 3.06 3.3 1.97 4.66

0.10

Estimated severe error rate. New York state regions outside of the Metropolitan New York City area.

b

Per fraction

0.18 (0) 3.22 0.29 0.25 0.32 (0.05 a)

0.037 (0.005 a)

Practical Radiation Oncology: January-March 2011

Some reports provide data regarding the types and causes of errors. The State of New York's data implicates “therapist error” in 84% of events, “failure to follow policies and procedures” in 63%, and “physics/dosimetry” in 27% (some attributed to multiple “causes”). A World Health Organization report reviewed significant events for more than 3 decades leading to death or injury in 3,125 patients.21 These errors were associated with treatment planning in ≈55%, the commissioning of new machines and systems in 25%, and data transfer and treatment delivery each in ≈9%. Among the 4,616 World Health Organization's events not leading patient harm, the identified sources of error were more diverse, included treatment planning in 9%, data transfer in 38%, treatment delivery in 18%, medical decision and assessment in 16%, simulation and imaging in 7%, and others in 28%. Reviews from Calgary and Radiation Oncology Safety Information System (a voluntary registry containing 1,200 events from 110, largely European, centers) implicated standards, procedures, and practices in ≈60%, communication in ≈15%, materials and equipment in ≈10%, and knowledge and skill in 10%.17 Thus, serious errors appear most associated with machine commissioning, dose calculation, treatment planning, and data transfer.21,22 Multiple reports note a shift in the type and frequency of errors with the introduction of new technologies21,23; eg, a decline in data entry errors with record and verify systems, an increase in “operator error” during equipment transitions when multiple machine types are used concurrently, and tasks are thus less standardized.10,23 High technology brings with it some automation, and a sense of infallibility, and increased detachment between the operator and the patient that may lead to more “low technology errors.”10,23 The inconsistencies in the data result from variable definitions, reporting requirements (eg, voluntary vs mandatory and anonymous vs regulatory), and confusion between types and causes of error. It is instructive to learn from our colleagues in human-factors engineering and psychology who study “human errors,” and from the fields of aviation and nuclear energy safety in which error reporting and analysis are more widely applied. These fields have largely overcome barriers to reporting (eg, cultures of nondisclosure due to shame, punitive reactions to reporting, fear of liability, and damage to professional reputation). Assessments of errors appreciate the lack of proximity between cause and effect, the diffused nature of responsibility, and the rarity of injury per error. Going forward, our field needs to embrace cultural changes to adopt a more systematic approach to reporting and understanding the causes of errors and harm.6,24-28 We need to develop clear criteria and definitions to categorize different types of errors, their cause(s), and so forth, to be able to facilitate analyses that lead to methods of prevention. It is particularly critical that this issue be addressed by the organizers of central data repositories

Safety in radiation oncology

5

for their pooled data to be most meaningful.16,29,30 For each event, when possible, there needs to be a clear distinction between what happened (or almost happened), when and who was involved, and what the multiple steps were that led to the event. Furthermore, it is important that we not waste time and effort arguing about nomenclature. The focus should not be on eliminating every error. Rather, we should focus our attention on errors that can cause real patient harm, and those that reflect systemic problems that might lead to errors more likely to cause harm.31

II. How did we get to this point? In the past, RT planning and delivery were less complex, and the pace of technology change was relatively modest. The time and work demands on personnel were fewer, and clinical procedures, with QA approaches, evolved with the technology. More recently, the pace of technology change has hastened, and our existing clinical procedures and QA approaches might be strained. Advanced treatment techniques often require vastly more information to be gathered, considered, analyzed, checked, and transferred, thus increasing the risks for error. In addition, the move from 2-dimensional (2D) to 3-dimensional (3D), and more recently to IMRT and image-guided radiation therapy (IGRT), and so forth, introduced several challenges, including the following (Fig 2).

1. Increased time demands and changing workflow Newer technologies require increased efforts for many members of the radiation oncology team (eg, for image segmentation, iterative dose calculations, patient-specific QA, image acquisition and review, treatment delivery, machine and multi-leaf collimator maintenance and repair). Individual's tasks are more interdependent, with more handoffs, thus increasing opportunities for delay and suboptimal information transfer (eg, dosimetrist image segmentation → medical doctor image segmentation and specification of dose and volume constraints → dosimetrist planning → medical doctor review → dosimetrist replan → iterate → and so forth). The time pressures on all involved are increased. The need for unambiguous communication and for easy-to-use tools is increased.

2. Addressing expectations Care providers, including referring physicians, patients, and administrators have been accustomed to our historic ability to proceed with consultation, simulation, and treatment initiation in rapid succession. With advanced

6

L.B. Marks et al

Practical Radiation Oncology: January-March 2011

Figure 2 (A) Confluence of events shown may have increased the risks of errors in radiation oncology. *The increased use of advanced technologies results, at least in part, from the physician's desire to provide state-of-the-art care, as well as patient, competitive, and financial pressures. (B) Sources of increased complexity can be somewhat arbitrarily divided into categories as shown, within radiation oncology, general oncology (eg, including issues related to chemotherapy and surgery), and broader issues affecting the whole health system. (C) Interventions aimed to reduce errors should consider the competing demands, concerns, and distractions that people face (eg, government regulations, educational missions, legal concerns, monetary and billing issues). (D) In concert, these many factors have increased demands to the point where they might exceed capacity. When there was excess capacity, minor inefficiencies in the workplace were more tolerable; workarounds and re-dos were relatively easy to accommodate. Presently, the calculus has changed to the point where demands are approached (or have already exceeded) capacity, leading to stress and an increased risk of error. Lean approaches will reduce the demands and increase the capacity (by reducing waste). DVHs, dose-volume histograms; GME, graduate medical education; HIPPA, health insurance portability and accountability act; IGRT, image-guided radation therapy; IMRT, intensity-modulated radiation therapy; MU, monitor units; QA, quality assurance.

technologies, this is less practical and might be dangerous. “Rushing” is a contributing factor in errors. The transition to 3D, particularly IMRT, lengthens the planning process. Administrators and providers need to be educated on the increased complexity of modern techniques and possible hazards of “work-arounds.” Similarly, patients and others must recognize that the increased capabilities of modern machines might lead to more frequent “down-times.” The instinct to move patients to different machines is, at best, logistically challenging, but it is often tremendously burdensome because some complex IMRT plans are machine-specific and re-planning and optimization and QA must be repeated. Despite these concerns, clinical care often dictates that we do rush a treatment plan, create

work-arounds, or re-plan patients (ie, all activities that stress the system and may lead to errors).

3. Beam-on time/monitor units/higher stakes of errors Multi-leaf collimator-based IMRT increases the monitor units per delivered dose. Catastrophic failures of the system (eg, multi-leaf collimators mis-positioned), although extremely rare, can have dramatic clinical consequences because the ratio between unintended and intended doses may be high. This was a factor, in one of The New York Times' reports.32 By comparison, failure

Practical Radiation Oncology: January-March 2011

to place a physical wedge typically has lesser, rarely catastrophic, dosimetric effects. Intensity-modulated radiation therapy markedly increases the number of parameters that need to be correctly generated and transferred, thus increasing opportunities for error.

4. The reliance on accurate image segmentation Three-dimensional and IMRT both rely on image segmentation. Institutions evolving from 2D to 3D to IMRT were able to refine their image-segmentation skills. Segmentation errors with 3D are often detected on digitally reconstructed radiographs or simulator and portal films (eg, the base of the tongue is challenging to segment on axial computed tomography, but its location is known on lateral planar images relative to the soft tissue and boney landmarks). Conversely, institutions rapidly moving from 2D to IMRT may have had less opportunity to refine image segmentation skills. Segmentation errors may be challenging to detect as they are not reflected in DVHs.

5. The reduced utility of some “end of the line” QA tools (Fig 1) a. Portal films verify isocenter location and treatment volume; these have been the mainstay of physician peer review for decades. Indeed, tumor borders are sometimes seen better on portal films than on simulator films (eg, lung periphery, pharyngeal). Errors in the planning and datatransfer process (eg, inadvertent rotation of a collimator), or changes in internal anatomy, are often detected on the portal films. Admittedly, portal images from unusually oriented beams are often not useful because the anatomic projections are less intuitively known. While IGRT verifies isocenter placement well, it does not replace the rapid global assessment of “treated volume” often afforded by portal films. With IMRT, beam orientations, apertures, and other parameters are less intuitive to the physician, and the ability to readily, albeit grossly, assess the treatment volumes is largely gone. Peer review, which was largely based on portal films, is now more challenging. b. “Checking the light field” on the patient's skin is a reasonable surrogate for the irradiated treatment volume with 2D- or 3D-planned beams. The light field allowed therapists to grossly assess if the correct volume was being treated (eg, if scars are adequately covered), if the field included critical normal tissues (eg, the eye), and was particularly helpful in considering match-lines and abutting fields. With IMRT, there is less connection between the light field and the irradiated volume, thus reducing the therapist's ability to assess the “reasonableness” of a treatment field. c. The number of monitor units (MUs) are less clearly related to delivered dose with IMRT versus nonIMRT techniques.

Safety in radiation oncology

7

6. Shorter treatment schedules Traditional RT is delivered during a span of many days and weeks. Thus, the dosimetric and clinical effects of a treatment deviation on a given day are diluted by the many other days of correct treatment delivery. Furthermore, a deviation on one day can often be “compensated for” on subsequent days. The move toward shorter and faster fractionation schedules reduces this ability to correct and compensate.

7. Tighter margins Traditional margins around a target considered multiple sources of uncertainty in concert, including target delineation, microscopic spread, target motion, and patient setup. Discrete target definitions (gross tumor volume, clinical target volume, internal target volume, planning target volume [PTV]) formally decouple these uncertainties (eg, clinical target volume addresses microscopic spread beyond the gross tumor volume, internal target volume addresses clinical target volume motion, and PTV broadly addresses setup inaccuracies). As technologies address these uncertainties (eg, gating and IGRT shrink internal target volume and PTV margins, respectively), there is a widely embraced, perceived ability to shrink overall margins (Fig 3). However, tumor control may suffer with tighter margins (Fig 3B). Concurrently, there are added complexities within the broader medical system (Fig 2B), and a growing number of increasing demands, concerns, distractions, and budgetary constraints on providers (Fig 2C). These many factors contribute to a general increased level of demands that cannot be met by available capacity (Fig 2D). For example, electronic medical records (EMR) greatly enhance care through things such as improved accessibility and legibility. However, the transition from a paper chart to EMR can be disruptive. Even when fully functional, it takes time to recreate workflows within EMRs. Some functions are diminished, multiple notes cannot be easily viewed concurrently or in rapid sequence, and thus the “clinical context” can often be difficult to appreciate. Notes typically cannot be annotated, and data entry and retrieval can be cumbersome. Some capabilities particularly helpful in our field can not be readily replicated in the EMR (eg, drawing and comparing serial pictures to document changes in tumor extent or normal tissue response). The lack of connectivity between many hospital EMRs and the (radiation) oncology-specific EMRs, presents particular concerns (eg, double data entry, unavailability of records to other departments, omissions and deletions). This may lead to stress and frustration and incomplete and erroneous documentation within the EMR, negatively affecting safety.

8

L.B. Marks et al

Practical Radiation Oncology: January-March 2011

Figure 3 (A) Margins are defined to compensate for microscopic spread, internal motion, and setup errors, as noted. Advances in imaging (eg, computed tomography [CT]) reduce uncertainty regarding gross tumor volume. Technology such as respiratory gating allows us to better control internal motion. Technologies, such as image-guided therapy, cone beam computed tomography, allow us to mitigate and reduce patient setup errors. Adaptation of these technologies led to a reduction in global margins, and in some instances, a decline in local control, perhaps due to the unmasking of uncertainties regarding microscopic spread. (B) As our certainty of gross anatomy increases, there is a tendency to reduce the field margin. However, there is probably a physical or biologically necessary margin related to our uncertainty of microscopic spread. If this uncertainty is not acknowledged, there is a chance of reducing the risk of margin leading to a marginal miss. Data from 2 studies demonstrate reduced local control with adoption of advanced approaches; IGRT for Engels and conformal therapy for Pfeffer. IGRT, image-guided radation therapy; PET, positron emission tomography.

In summary, the increased concerns for safety results, in large part, from the adoption of advanced technologies and associated commissioning, dose calculation, and data transfer. Safety concerns also stem from fundamental

changes in workflow (eg, multiple hand offs), the relative loss of some traditional “end-of-the-line” QA tools (port films and light fields), changes in fractionation schedules, and perhaps an under-appreciation for the physical

Practical Radiation Oncology: January-March 2011

Safety in radiation oncology

9

limitations of imaging-based diagnosis and treatment. Technology can also promote complacency (ie, “it must be right, the computer said so”). Furthermore, there are other broad issues within health care (eg, regulatory, insurance) that strain our systems and may influence safety. Change is a major source of risk. As stated by Youngberg and Hatlie,33 “… change is creating new paths for failure and new demands on worker… revising their understanding of these paths is an important aspect of work on safety; … missing the side effects of change is the most common form of failure for organizations and individuals.” Measures to enhance safety should recognize the multi-faceted nature of changes in our field and the associated challenges.

III. Opportunities to enhance safety The confluence of events previously noted (eg, Fig 2) have strained existing workflow and QA systems, and have likely led to some negative clinical events. The vast majority of patients treated with traditional and advanced technologies are treated safely as intended. Much of the effort to enhance safety focuses on the tangible (eg, hardware and software). These efforts are certainly necessary. However, we often fail to acknowledge that the “user/operator” has variable abilities and training, and performs under various operating conditions, equipment configurations, and work scenarios. Only the total composite of these elements, and the human component, will determine the safety of the system. Several broad concepts that can be applied to enhance safety are discussed.

1. Staffing/schedules/facilities Staffing levels need to be adjusted to reflect the workload, particularly in physics and dosimetry where the demands have markedly increased (eg, patient-specific QA for IMRT). Systems that facilitate clear, unambiguous, and efficient communication between physicians and dosimetrists/physicists are critical, given the large number of hand-offs and interdependent tasks. This might require locating planning areas closer, even within the clinic offices and/or dedicating time for physicians and dosimetrists to work together, to facilitate the iterative “segment-plan-review-repeat” cycle. It might be impractical for physicians and planners to rotate between facilities, and if they must, software should enable efficient and accurate communication/transfer of complex 3D data between centers.

2. Workflow/efficiency/standardization (Fig 4) Clinical practice is complex, often mired in administrative and historically derived procedures. Efficiency impacts quality and safety. Hassled workers are more prone to error,

Figure 4 (A) Tasks may be considered a continuum from simple to difficult (x-axis). The operators' control and ability are shown on the y-axis. Highly complicated systems operated by extremely functional individuals are most effective and yield the skilled operator much satisfaction. Complex systems in which the operator lacks either control or abilities are the source for anxiety, stress, and an increased likelihood of error. (B) Standardization and usability improvements lead to more intuitive systems and a reduction in associated risks.

and the removal of waste will increase time available for critical tasks. We and others34,35 are applying lean approaches (adapted from the Toyota Production System)36 to streamline clinical workflow and alter the work environment (Fig 2D). Using rapid improvement events (Kaizens), participating representative members of involved groups create process maps for particular tasks. For example, at UNC, a physician, nurse, scheduler, simulator therapist and dosimetrist came together to address the scheduling of a patient for computed tomographic simulation (“CT-sim”) (ie, computed tomographic imaging for radiation therapy planning). Value-added steps are identified, with wasteful steps and unnecessary stressors being eliminated, and a more streamlined, unambiguous, standardized process being defined. Having stakeholders meet to discuss and define their work builds teamwork, mutual respect, and it also fosters an environment in which people know that they can positively impact their work. Standardization of clinical workflow also improves quality. Using our scheduling of CT-Sim example, our Kaizen was needed because too many patients were doublebooked, coming for simulation unprepared (eg, without oral contrast), without consent, and without necessary setup information, leading to delays and rescheduling. During the Kaizen, members identified problems, such as too many different individuals scheduling CT-Sim, some without full knowledge of the process, and the lack of an unambiguous means for physicians to communicate CT-Sim instructions. For scheduling, they now agreed to use a single mobile

10

L.B. Marks et al

phone number answered by a Sim therapist. The therapist is the only person permitted to schedule a patient and determines that the patient has all the necessary instructions. In addition, we created a single location within Mosaiq (IMPAC Medical Systems, Inc., Sunnyvale, CA) for physicians to enter simulation instructions. This has eliminated 95% of the related delays and need for rescheduling patients. Similarly, standardization is useful to avoid confusion during the planning process. Beams can be given clear, unambiguous names that describe their orientation,37 target and creator (eg, A15L_Rt Supraclav_LM is a beam from the anterior, angled 15° degrees to the left, directed at the right supraclavicular area created by LM). This is more descriptive than “LAO”. A similar concept applies to treatment plans. In group practices, it is helpful if providers can agree on standard approaches to common diseases using reference or guide sheets to avoid confusion among planning staff. Pre-printed labels with unique identifiers and barcodes are a reasonable way to facilitate uniform information transfer, reduce errors in spelling and data entry, link patients to their records, and identify patients. At UNC, patients have a card with a unique barcode that they scan at each treatment visit. This “self-check-in” cues the patient to the schedule of the electronic treatment delivery system. In addition, the patient can see his or her name appear on a screen associated with a particular treatment machine. If there is an error, the patient is empowered to alert the nearby check-in staffer, involving the patient in their care.

Practical Radiation Oncology: January-March 2011

3. Hierarchy of effectiveness 38 Different methods used to affect behaviors have variable expectations for success (Fig 5). Reliance on policies and training is our usual, but least effective, approach. In the New York state database, “failure to follow policies/ procedures” was implicated as a contributing factor in 84% of events, versus “inadequate policies/procedures” in 16% of events. Whenever possible, it is best to “hardwire” the systems for success using simplification, standardization, automation, and forced functions to create workflows and systems that support human work. Checklists and timeouts are effective,39,40 especially if they are focused, if the user believes in their utility, or if they are forced to use them, or both (eg, a “hard stop”). “Knowledge in the field” (automatic computer and machine functions, and checklists) is more likely to improve human performance than is “knowledge in the head” (memory). For the example previously given, at UNC we have modified our treatment planning system, such that all saved treatment plans are given a default name that includes the date, time, and name of the person who is logged into the computer. In this case, the user often appends some additional descriptive information (eg, treatment site). We have found this to be helpful in sorting through the directories with multiple plans.

4. Human factors engineering 38,41 Human-machine interactions are ubiquitous. Human factors engineering aims to define processes, interfaces,

Figure 5 Hierarchy of effectiveness.38 Different approaches to modify behavior have difference expectations for success, as shown on the left. Shown on the right is a real example from radiation oncology, and the goal is to have a uniform approach to naming the treatment plans. Computer-based automatic naming of plans forces the desired behavior, thus providing a compliance rate beyond that which is achievable with education and policies.

Practical Radiation Oncology: January-March 2011

Safety in radiation oncology

11

Figure 6 There are ways to design human-machine interfaces to facilitate their correct usage. Failure to do this in the workplace will increase the risks of error. There are “generic opportunities,” such as in the design of the keypads in phones versus computer keyboards (panel A). Presently, the numbers are not placed in a consistent fashion, perhaps leading to data entry errors. Panel B is an example from our radiation oncology-specific record-and-verify system. The word “status” is used to describe if a beam's prescription is approved or disapproved. The status of a single beam (left chest wall-thin dark oval on the left) is shown in the center (thick solid dark oval). The status of that beam can be changed by clicking on the word “status” in the lower right hand corner (white solid circle). However, the status of multiple other beams (if they are “tagged” with the symbol [N] in the upper left portion of the screen) may also be modified concurrently. Note that the vertical height of the upper portion of the screen, where the different beams are listed, is limited, and a scrolling function is provided to allow the full listing of beams to be viewed (oval upper right). Because all of the beams may not be visible concurrently, depending on the number of beams, the user is able to “approve or disapprove” beams whose names are not visible on the screen (ie, a tagged beam may not be seen if the list of beams is long). The word status is also used in the lower center of the screen (dotted dark oval), but this has nothing to do with beam approval or disapproval. The ovals and arrows in the display were added to facilitate explanation.

and machinery that facilitate correct usage. For example, the forcing function of an automated teller machine can require withdrawal of the bankcard before money is

dispensed. Similarly, placing console control buttons that perform particular functions in a consistent location enables users to more reliably operate equipment in a

12

L.B. Marks et al

predictable and correct manner. Safety is improved with workspaces that are designed to reduce noise, interruptions, and visual clutter while improving lighting, temperature, desk height, and other factors proven to affect performance.38 In our field, complicated computer screen layouts, keyboard functions, and treatment consoles are a few examples of the many of human-machine interfaces that we navigate daily. These may require increasing mental effort as they become more complicated or lack standardization (Fig 6). Many are well designed, but there is ample room for improvement. For example, within individual products, short-cut keyboard commands should be consistent whenever possible.

5. Incorporating QA tools/functionality into software Often, QA is not incorporated into many of our field's planning or record-and-verifying delivery systems. For example, user-configurable checklists and time-outs are not an option. Although potentially valuable, such imbedded checklists still require the user to verify that checklist items are appropriately addressed rather than being automatic. Imbedded automatic QA functions would be useful, such as: a. For a new plan, the system searches its directory archive for patients with the same name to identify inadvertent retreatment. b. For common diagnoses, the planning system compares the proposed target volumes and associated dose parameters to a library of userspecified “expected” parameters and issues predefined alerts. c. Normal tissue dose-volume parameters are compared to user-specified constraints. d. Automatic highlighting of under-dosed target, or normal tissue hot-spots. e. Beams and plans are named automatically to reflect the treatment planner, date, and so forth. Some of these functions may already exist. McNutt is “training” their planning system to identify discrepancies between pending plans and their library of “similar plans.”42 We have had beam auto-naming for years,37 and recently we added automatic naming of saved treatment plans.

6. Reassessing peer review/daily morning meetings Traditional physician peer review (ie, after initiating RT) is less useful with IMRT. The threshold to making a meaningful change in image segmentation is relatively high because it may result in time-consuming re-planning and QA. More subtle peer-recommended changes are likely unaddressed if the real or perceived effort-toimplement is “high.” Conducting physician peer-review

Practical Radiation Oncology: January-March 2011

prior to planning may be a more effective way to conduct physician peer review. At UNC, we have a daily morning meeting at which we discuss new patients who have had their CT-Sim and are going to be planned. This is widely attended by residents, mid-levels, attendings, physics and dosimetry staff, senior therapists, and dosimetry and therapy students. Others are welcome, and often attend. We review the indications for treatment, image-segmentation, and tentative doses. Sometimes tentative beams and plans have already been generated (eg, simple fields and non-IMRT cases). The rate of changes in image segmentation or planned treated volume is higher in our pre-RT peer review than in traditional “chart rounds” review conducted after treatment has started.43 In this meeting, dosimetrists routinely note inconsistencies in the segmentations and directives, and anticipate dosimetric challenges (eg, “I cannot meet both the cord and the PTV doses due to their proximity”). Physicians and dosimetrists engage in a healthy dialogue that we believe makes the subsequent planning process smoother. “Huddle.” The morning meeting includes a “Huddle” to review that day's pending activities, including: a. CT-Sim Schedule: The therapists note patients whose records lack clear directives. Patients presenting unique challenges or learning opportunities are noted. The availability, or lack, of openings for add-ons is noted. b. Dosimetrists alert the group regarding treatment plans that are proceeding more slowly than expected and seek direction. c. The daily patient treatment census is noted along with special considerations that might affect clinical flow (eg, anesthesiology or total body irradiation cases). d. Announcement is made of patients who will need pre-RT films so that the responsible physician(s) can plan to be available. e. The name of the “Doc of the Day” is announced (ie, the physician responsible for inpatient coverage and unanticipated clinical needs). f. The group is invited to raise concerns, make announcements, and so forth. The morning meeting serves the practical functions of trying to anticipate the upcoming challenges and avoid chaos in the clinic. It also serves a social and cultural function to bring the department together daily, fostering an environment of easy communication among all team members.

7. “Safety rounds” Frankel et al44 described components of high reliability organizations, as including those with “mindfulness,” high alertness for potential mistakes, use of assertive language, having responsive leadership, and a safe culture for

Practical Radiation Oncology: January-March 2011

reporting errors and near misses. In 2010, the department at UNC instituted leadership “Safety Rounds” characterized by personal, 15- to 20-minute interviews by the chairman and members of the leadership team with staff members in groups of 1 to 3 people at their work site, asking about near-misses or unsafe conditions causing potential or real harm to patients or employees.

8. Fostering a “culture of safety” These many activities help foster a sense of openness, mutual respect, group participation, and responsibility. We believe this is particularly important as a recent survey suggests that 31% of therapists are uncomfortable speaking with their physician colleagues regarding safety concerns if a mistake has been made.45 Staff should be encouraged to raise concerns and be reassured that reporting is not punitive. Honest errors occur and can be forgiven; however, people need to be held accountable for intentional deviations from safe practice. This section is not intended to be exhaustive. Our focus has been primarily on external beam RT (eg, implants and intraoperative RT were not considered). There are many other societal-wide initiatives that are opportunities to enhance safety. For example, Integrating the Healthcare Enterprise-Radiation Oncology is a long-standing initiative within ASTRO/AAPM to better facilitate connectively between products from different vendors.46-48 Creating such cross-vendor electronic connections is often frustrating and time consuming. The ASTRO is sponsoring the creation of multiple “white papers” to raise awareness and address many safety issues in our field, specifically as they relate to IMRT, IGRT, stereotactic body radiation therapy, and brachytherapy. These papers will include much useful information and suggestions regarding workflow and QA.

IV. Summary RT is an effective treatment providing the opportunity for cure and palliation to millions of patients annually. Several evolving changes in our field may have increased the risks of error. There are enormous efforts being expended nationally, by many people and organizations, to address these issues. The highly technical nature of our field requires that safety initiatives address these RT-specific technical factors (eg, defining standards for complex procedures, such as IMRT). However, it is important that we recognize the broader challenges that influence our practice, and that we learn from our colleagues in other industries, such as aviation. Defining better policies is not enough. Standardization, automation, recognition of the “hierarchy of effectiveness,” defining unambiguous means of communication, checklists, and

Safety in radiation oncology

13

consideration of “human factors” in the design of products and workspaces, are all powerful tools to improve safety. It is critical that all members of the team understand these concepts, and that they be involved in maintaining safety. Leaders need to create a culture of safety and open dialogue,49 through initiatives such as safety rounds, the morning huddle, and Kaizen events. Ongoing changes in our clinical practice mandate continued vigilance to minimize the risks of error.

Acknowledgment Thanks to Suzanne Bodeen for assistance in preparation of this manuscript, and to the many members of the Department of Radiation Oncology, UNC Health System Administration, and the NC State Industrial Extension Service, that participated in, facilitated, and supported many of the activities described.

Appendix I American Society for Radiation Oncology's Target Safety Plan: 1. Work to create a national database for the reporting of medical errors. 2. Advocate for new and expanded federal initiatives to help protect patients from radiation errors; support the immediate passage of the Consistency, Accuracy, Responsibility and Excellence in Medical Imaging and Radiation Therapy (“CARE”) Act, which among other things requires national standards for radiation therapy treatment team members. 3. Work with cancer support organizations to help cancer patients and their families know what to ask their doctors when radiation is a possible treatment option. 4. Enhance the radiation oncology practice accreditation program, and develop additional accreditation classes specifically addressing new technologies. 5. Expand our educational training programs to include an intensive focus on quality assurance and safety. 6. Accelerate an ongoing effort that seeks to ensure device manufacturers can transfer treatment information from one machine to another seamlessly to reduce the chance of a medical error.

References 1. Bogdanich W. Safety features planned for radiation machines. N Y Times. 2010;A19. 2. Bogdanich W. VA is fined over errors in radiation at hospital. N Y Times. 2010;A20. 3. Bogdanich W, Ruiz RR. Radiation errors reported in Missouri. N Y Times. 2010;A17.

14

L.B. Marks et al

4. Bogdanich W. Radiation offers new cures, and ways to do harm. N Y Times. 2010;A1. 5. Goldwein JW, Podmaniczky KC, Macklis RM. Radiotherapeutic errors and computerized record/verify systems. Int J Radiat Oncol Biol Phys. 2003;57:1510. 6. Fraass BA, Lash KL, Matrone GM, et al. The impact of treatment complexity and computer-control delivery technology on treatment delivery errors. Int J Radiat Oncol Biol Phys. 1998;42:651-659. 7. French J. Treatment errors in radiation therapy. Radiat Ther. 2002;11:149-159. 8. Huang G, Medlam G, Lee J, et al. Error in the delivery of radiation therapy: results of a quality assurance review. Int J Radiat Oncol Biol Phys. 2005;61:1590-1595. 9. Macklis RM, Meier T, Weinhous MS. Error rates in clinical radiotherapy. J Clin Oncol. 1998;16:551-556. 10. Marks LB, Light KL, Hubbs JL, et al. The impact of advanced technologies on treatment deviations in radiation treatment delivery. Int J Radiat Oncol Biol Phys. 2007;69:1579-1586. 11. Patton GA, Gaffney DK, Moeller JH. Facilitation of radiotherapeutic error by computerized record and verify systems. Int J Radiat Oncol Biol Phys. 2003;56:50-57. 12. Yeung TK, Bortolotto K, Cosby S, et al. Quality assurance in radiotherapy: evaluation of errors and incidents recorded over a 10 year period. Radiother Oncol. 2005;74:283-291. 13. Barthelemy-Brichant N, Sabatier J, Dewe W, et al. Evaluation of frequency and type of errors detected by a computerized record and verify system during radiation treatment. Radiother Oncol. 1999;53: 149-154. 14. Ford EC, Terezakis S. How safe is safe? Risk in radiotherapy. Int J Radiat Oncol Biol Phys. 2010;78:321-322. 15. Williamson JF, Thomadsen BR. Foreword. Symposium “Quality assurance of radiation therapy: the challenges of advanced technologies”. Int J Radiat Oncol Biol Phys. 2008;71:S1. 16. Williams TR. Target safely. Available at: http://www.huffington post.com/tim-r-williams-md/target-safely_b_544270.html. Accessed August 4, 2010. 17. Dunscombe P. What can go wrong in radiation treatment? Presented at the AAPM/ASTRO Safety In Radiation Therapy Meeting; Miami, FL: June 24-25; 2010. 18. Ibbott GS. What can go wrong in radiation treatment: data from the RPC. Presented at the AAPM/ASTRO Safety in Radiation Therapy Meeting; Miami, FL: June 24-25; 2010. 19. Dansereau RE. Misadministrations-event summaries and prevention strategies. Troy, NY: State of New York Department of Health; 2010. BERP 2010-1. 20. Donaldson SR. Towards safer radiotherapy. London: British Institute of Radiology, Institute of Physics and Engineering in Medicine, National Patient Safety Agency, Society and College of Radiographers, The Royal College of Radiologists; 2007. 21. World Health Organization. Radiotherapy risk profile: technical manual. Geneva, Switzerland: WHO Publishing; 2008. 22. Preventing accidental exposures from new external beam radiation therapy technologies. International Commission of Radiological Protection; 2010. 23. Amols HI. New technologies in radiation therapy: ensuring patient safety, radiation safety and regulatory issues in radiation oncology. Health Phys. 2008;95:658-665. 24. Ford EC, Gaudette R, Myers L, et al. Evaluation of safety in a radiation oncology setting using failure mode and effects analysis. Int J Radiat Oncol Biol Phys. 2009;74:852-858. 25. Pawlicki T, Mundt AJ. Quality in radiation oncology. Med Phys. 2007;34:1529-1534. 26. Pawlicki T, Whitaker M. Variation and control of process behavior. Int J Radiat Oncol Biol Phys. 2008;71:S210-S214. 27. Pawlicki T, Whitaker M, Boyer AL. Statistical process control for radiotherapy quality assurance. Med Phys. 2005;32:2777-2786.

Practical Radiation Oncology: January-March 2011 28. El Bardissi AW, Wiegmann DA, Dearani JA, et al. Application of the human factors analysis and classification system methodology to the cardiovascular surgery operating room. Ann Thorac Surg. 2007;83: 1412-1419. 29. IAEA. Recently raised issues of radiation risks to patients. Available at: http://rpop.iaea.org/RPOP/RPoP/Content/News/recent-issues-ra diation-risks.htm. Accessed September 20, 2010. 30. ROSIS. Radiation Oncology Safety Information System. Available at: http://www.clin.radfys.lu.se/default.asp. Accessed September 20, 2010. 31. Woods DD, Cook RI. Mistaking error. In: Youngberg BJ, Hatlie MJ, editors. The Patient Safety Handbook. Sudbury, MA: Jones and Bartlett; 2004:95-108. 32. Bogdanich W. As technology surges, radiation safeguards lag. N Y Times 2010;A1. 33. Youngberg BJ, Hatlie MJ. The Patient Safety Handbook. Sudbury, MA: Jones and Bartlett; 2004. 34. Kim CS, Hayman JA, Billi JE, et al. The application of lean thinking to the care of patients with bone and brain metastasis with radiation therapy. J Oncol Pract. 2007;3:189-193. 35. Mazzocato P, Savage C, Brommels M, et al. Lean thinking in healthcare: a realist review of the literature. Qual Saf Health Care. 2010;19:376-382. 36. Going Lean in Health Care. Secondary going lean in health care. Cambridge, MA: Institute for Healthcare Improvement; 2005. Available at: www.IHI.org. 37. Sailer SL, Bourland JD. Rosenman JG, et al. 3-D beams need 3-D names. Int J Radiat Oncol Biol Phys. 1990;19:797-798. 38. Cafazzo J. Healthcare human factors. Presented at the AAPM/ ASTRO Safety in Radiation Therapy Meeting; Miami, FL: June 24-25; 2010. 39. Gawande A. The Checklist Manifesto: How to Get Things Right. 1st ed. New York: Metropolitan Books; 2010. 40. Marks LB, Hubbs JL, Light KL, et al. Improving safety for patients receiving radiotherapy: the successful application of quality assurance initiatives[Abstract]. Int J Radiat Oncol Biol Phys. 2008;72:(suppl 1) S143. 41. Rivera AJ, Karsh BT. Human factors and systems engineering approach to patient safety for radiotherapy. Int J Radiat Oncol Biol Phys. 2008;71:S174-S177. 42. Wu B, Ricchetti F, Sanguineti G, et al. Patient geometry-driven information retrieval for IMRT treatment plan quality control. Med Phys. 2009;36:5497-5505. 43. Adams R, Church J, Chang S, et al. Quality assurance in clinical radiation therapy: a quantitative assessment of the utility of peer review in a multi-physician academic practice. Presented at the American Society for Therapeutic Radiology and Oncology (ASTRO); Chicago, IL: October 31-November 5; 2009. 44. Frankel A, Grillo SP, Pittman M, et al. Revealing and resolving patient safety defects: the impact of leadership WalkRounds on frontline caregiver assessments of patient safety. Health Serv Res. 2008;43: 2050-2066. 45. Adams R, Chen R, Bailey J, et al. The willingness of dosimetrists and radiation therapists to challenge physicians as a means of assuring patient safety: result of a pilot program. Presented at the American Society for Therapeutic Radiology and Oncology (ASTRO); San Diego, CA: October 31- November 4; 2010. 46. IHE Radiation Oncology. Available at: http://wiki.ihe.net/index.php? title=Radiation_Oncology. Accessed September 22, 2010. 47. Abdel-Wahab M, Rengan R, Curran B, et al. Integrating the healthcare enterprise in radiation oncology plug and play–the future of radiation oncology? Int J Radiat Oncol Biol Phys. 2010;76:333-336. 48. Tripuraneni P. Accreditation. Presented at the AAPM/ASTRO Safety in Radiation Therapy Meeting; Miami, FL: June 24-25; 2010. 49. Pronovost PJ, Vohr E. Safe Patients, Smart Hospitals: How One Doctor's Checklist Can Help Us Change Health Care From the Inside Out. New York: Hudson Street Press; 2010. 50. Errors in radiation therapy. Pa Patient Saf Advis. 2009;6(3):87-92.