The decision aid is the easy part: workflow challenges of shared decision making in cancer care

Abstract

Delivering high-quality, patient-centered cancer care remains a challenge. Both the National Academy of Medicine and the American Society of Clinical Oncology recommend shared decision making to improve patient-centered care, but widespread adoption of shared decision making into clinical care has been limited. Shared decision making is a process in which a patient and the patient’s health-care professional weigh the risks and benefits of different options and come to a joint decision on the best course of action for that patient on the basis of their values, preferences, and goals for care. Patients who engage in shared decision making report higher quality of care, whereas patients who are less involved in these decisions have statistically significantly higher decisional regret and are less satisfied. Decision aids can improve shared decision making—for example, by eliciting patient values and preferences that can then be shared with clinicians and by providing patients with information that may influence their decisions. However, integrating decision aids into the workflows of routine care is challenging. In this commentary, we explore 3 workflow-related barriers to shared decision making: the who, when, and how of decision aid implementation in clinical practice. We introduce readers to human factors engineering and demonstrate its potential value to decision aid design through a case study of breast cancer surgical treatment decision making. By better employing the methods and principles of human factors engineering, we can improve decision aid integration, shared decision making, and ultimately patient-centered cancer outcomes.

Patient-centered cancer care and human factors engineering

Almost 2 million new cancers will be diagnosed in the United States in 2023 (1). Preference-sensitive decisions—those in which evidence supports more than 1 management option, whereby choices must rest upon patient preferences and values—are found across the cancer care continuum, from genetic testing and screening (eg, lung cancer screening with low-dose computed tomography scanning) to treatment following diagnosis (eg, lumpectomy vs mastectomy in early-stage breast cancer) to posttreatment surveillance methods (2,3). The optimal approach to ensure that these complex decisions are patient centered is shared decision making, an interactional process where a health care professional and a patient use the best available evidence and information on patient values, preferences, and goals to reach medical decisions together (4). Patients with cancer who are less involved in this decision-making process are statistically significantly less satisfied and have higher decisional regret (5), whereas patients who engage in shared decision making report higher quality of care and more positive perceptions of physician communication (6,7).

Challenges in delivering genuinely patient-centered care have been recognized for decades (2,8-11), and shared decision making remains limited in clinical practice (12,13). For instance, a key step in shared decision making is providing patients with relevant information and eliciting their values, concerns, and goals (14). However, fewer than half of patients believe that their treatment reflects their preferences (15,16) and levels of patient involvement in decision making remain low (17).

In this commentary, we argue that many of the most challenging barriers to achieving effective shared decision making in cancer care can be grouped under the umbrella of workflow. Workflow describes how people, tasks, tools, technologies, and information interact over time and space in the service of work goals (18). Analyzing and optimizing workflow are necessary to ensure that shared decision making is effective, efficient, and equitable. Furthermore, to support shared decision making, these analyses must accommodate the difficult fact that care workflows are often complex, nonlinear, and multimodal (19,20), conducted by clinicians, patients, and their caregivers, with input from many other members of the cancer care team. Fortunately, the discipline of human factors engineering offers methods for understanding and improving complex workflows. We argue that a human factors engineering perspective is needed to achieve patient-centered decision making in cancer care and show how such a perspective can inform better ways to implement shared decision making—for example, by employing a human factors engineering methodology called human-centered design to improve the design of decision aids.

Promoting shared decision making through decision aids

For several decades, a focus of shared decision-making research in oncology care has been the development of decision support tools (or decision aids) for patients (16,21,22). Decision aids should contain unbiased and structured information about possible options, with the risks and benefits of each option, and should support patient reflection on their own values and preferences (23,24). Decision aids can be videos that patients view at home (25), paper forms used in conversations with clinicians (26), apps on patients’ phones, or web-based portals (27). Decision aids may be particularly helpful for making cancer treatment decisions given the growing number of viable options between which patients must decide, each with differing impacts on patients’ short-term and long-term quality of life (28). Further, cancer treatment decisions are complicated by emotional (eg, anxiety, shock) and cognitive (eg, limited understanding of medical and risk information) factors. Decision aids can improve patients’ knowledge of their options, improve their expectations of different options’ potential benefits and harms, reduce decisional conflict, improve patient satisfaction with the decision-making process, and reduce the number of unnecessary surgical treatments (9,29,30). Yet, to date, patient-facing decision aids have had limited impact on cancer care.

Workflow barriers to integrating decision aids into cancer care

Despite the strong evidence in favor of decision aids and the plethora of decision aids available for cancer decisions, study after study has identified serious workflow-related barriers to implementing them in routine care (11,31-34). In these studies, clinicians expressed concern that decision aids would disrupt clinical workflows, stating that there was insufficient time to use them, and disagreeing on their optimal content and format. The Healthcare Information and Management Systems Society has outlined 5 rights of decision support: The support should convey the right information to the right person in the right format through the right channel and at the right point in the workflow (35). In the context of decision support for shared decision making, the 5 rights must be considered not only for the patient and the clinician individually but also for joint patient-clinician interactions. For instance, what aspects of decision aid design and delivery are needed for the patient, for the clinicians, and for their interaction together? What is the optimal format of the decision support for these different roles? What channel should be used to deliver the decision aid to the various parties (eg, electronic health record, paper, video, phone call, text message)? Perhaps most challenging, at what point in the workflow should the decision aid be delivered to patients and clinicians? Here, we highlight 3 specific workflow-related issues encountered in the implementation and use of decision aids for cancer care:

• Who should use the decision aid?

• When is the optimal time to use the decision aid?

• How do existing clinical workflows support or hinder shared decision making?

Who should use the decision aid?

Identifying the intended users of the decision aid can help ensure that it is designed to meet the users’ workflow, goals, and needs. Shared decision making is understood to require the involvement of the patient and at least 1 clinical expert. As a result, at least 2 versions (ie, user interfaces) of a decision aid may be required. For example, tools developed for clinicians are likely to include more technical language and assume deep background medical knowledge, whereas tools for patients may use simpler terms and provide more definitions, explanations, and illustrations of the medical or statistical concepts (36). It is not yet known whether and when such decision aids should be used by the patient and the clinical expert separately or together in a joint discussion. It is possible that a third version of the decision aid would better support the joint discussion and shared decision making.

In addition, it may not be clear exactly who should be the clinician users and the facilitators of patient use. In a randomized controlled trial of a decision aid on adjuvant breast cancer therapy (37), trained health educators presented the decision aid and associated information. How many cancer clinics have health educators on staff? Alternatively, which other care team members might assume this responsibility? Will they be able to do it as effectively? The Creating better Health Outcomes by Improving Communication about patients’ Experiences (CHOICE) decision aid (38), which supported patient-centered cancer symptom management, was intended for use by both physicians and nurses. In a survey of clinicians, however, Ruland (38) found that nurses used the information from CHOICE statistically significantly more often than physicians and that nurses rated the usefulness of CHOICE higher than physicians. Légaré et al. (13) acknowledged the numerous health-care roles involved in shared decision making and described the need to expand shared decision-making research beyond the “dyad-based approach, one health professional interacting with one patient.” Identifying the right stakeholders will help ensure that we can design systems that deliver the right information in the right format (35).

The role of patient family caregivers is another challenge to decision aid design. For instance, not all patients have family caregivers, and caregivers can have differing relationships with patients (eg, neighbor, husband, adult child) and differing levels of desired involvement in decision making. Shared decision-making workflows and decision aids need to support the variable nature of caregiver involvement in the decision-making process.

When is the optimal time to use the decision aid?

Another challenge relates to the timing of the delivery of decision aids to patients or clinicians. In the 5 rights framework, we need to ensure that the information is delivered at the right point in the workflow (35). In a randomized trial, a decision aid about genetic testing for risk of breast and ovarian cancer statistically significantly improved knowledge but did not affect actual choices or reduce decisional regret (39). The researchers concluded that suboptimal timing of the decision aid may have limited its effectiveness. For instance, women who had already given blood (which is used for genetic testing) on the day they received the decision aid were less likely to benefit, possibly because the genetic testing decision may have already been made. In a study of a text-based and picture-based decision aid to support surgeon-patient conversations about breast cancer treatment (40), surgeons disagreed on the optimal timing to deliver the decision aid, with half stating that it should be given to patients before the clinic visit and the other half suggesting it should be provided during the visit with the surgeon. During development of a decision aid for women with advanced breast cancer (41), patients expressed concerns that receiving the tool right after diagnosis would be overwhelming. Patients also expressed a desire to use the tool before appointments, however, to foster greater consideration of their values and preferences. The researchers concluded that more research is needed to determine the appropriate time to give the decision aid to patients.

In addition to finding the right time in the workflow to access a decision aid, there are challenges with having sufficient time to use the aid before a decision is needed. In a multisite study of the implementation of a decision aid for early-stage breast cancer surgical treatment, Silvia and Sepucha (25) explained, “Patients need about an hour to view the programme and they need to view it at the right time, after diagnosis and before they choose what type of surgery to have.” Many sites reported time constraints and challenges getting decision supports to patients before they made a surgical decision. One clinic developed a specific process to integrate the decision aid into patient flow; patients received an “appointment” to view the decision aid before meeting with a surgeon, which supported more effective use of the tool. Hahlweg et al. (42) implemented 2 decision aids to support breast cancer surgical decisions and breast reconstruction following mastectomy, finding that the decision aids were used in only one-third of eligible patient encounters. The right time to use the decision aid was controversial among physicians, and time pressure was also reported as a barrier to use. Several studies have faced similar challenges getting decision aids to patients before the decision is made and have described insufficient time as a barrier (43,44).

Although systematic reviews have identified time constraints as a barrier to decision aid use (33,43,44), evidence is limited for whether decision aids actually add time to visits (33). For instance, Whelan et al. (2003) (45) did not find statistically significant differences in consultation time between visits with and without decision aid use. It is also unclear whether clinicians’ perception of insufficient time to use decision aids is related to actual time required or inadequate incentives (eg, financial, clear benefit to patients) to use the tools. Most likely, it is a combination of these factors: Decision aids take too much time to use because they are frequently not designed to be quick and easy to use, they are not seamlessly integrated in clinical workflows, and physicians are not motivated to overcome these barriers to use the tools because there is no clear incentive to do so. These challenges must be addressed to improve decision aid design and use.

How do existing clinical workflows support or hinder shared decision making?

A third challenge to the use of decision aids is erroneous assumptions about or limited understanding of how existing workflows can influence decision aid use. In a randomized controlled trial of an electronic health record–embedded decision aid about prostate cancer (46), only 20% of eligible men actually received the decision aid. The researchers had assumed that patients would be referred to the decision aid during the telephone call informing them of their diagnosis before their clinic visit. However, most patients were never called before the clinic visit, and therefore did not receive the decision aid as intended. In a study of the CHOICE decision aid designed to support cancer symptom management (38), nurses described challenges integrating the tool into daily routines. Even after a year of use, a nurse explained, “Maybe part of it is that CHOICE is fairly new, and there are so many familiar old routines you sort of have to rearrange if you’re going to make this a good part of those routines.” Nurses experienced challenges accessing CHOICE because several different systems were used for planning patient care, forcing nurses to log in to a different system to access the decision aid. They also reported that some patients thought the system was cumbersome to use. Schubbe et al. (40) developed and implemented 2 decision aids for early-stage breast cancer to be used during conversations with surgeons as part of a randomized controlled trial. In postuse interviews, half the surgeons thought that the decision aid disrupted the flow of their conversations with patients. They also reported that the information in the decision aid was not in the correct order. In contrast, Salwei et al. (47) applied human factors engineering methods to analyze workflow patterns of a decision aid in the emergency department, identifying potential barriers to use and design modifications needed before implementation. In a scenario-based simulation, the decision aid was found to reduce workload and improve diagnostic accuracy compared with the current system and was preferred by 88% of physicians (48,49). In the sections that follow, we describe how human factors engineering methods can be used to address workflow challenges and facilitate integration of decision aids into routine cancer care.

How can we do this more effectively? The role of human factors engineering

Human factors engineering is the scientific discipline aimed at understanding the interactions between humans and other system elements, and the profession that applies methods, principles, and theory to design with the goal of optimizing human well-being and system performance (50). Human factors engineers are trained to apply a holistic perspective to study complex sociotechnical systems, focusing on physical, cognitive, and organizational factors in work. To do this, they engage multidisciplinary teams and use both qualitative and quantitative methods. More than 100 human factors engineering tools for analyzing workflow have been described (18). A framework and checklist have also been developed to support the integration of technologies into clinical workflows (47,51). These approaches can support workflow analysis across all stages of decision aid development, from work system analysis and design to implementation and postdeployment improvement.

Human-centered design

Human-centered design (ie, user-centered design) is one human factors engineering method to support the development of tools that meet the needs of users (52). Figure 1 shows the 5 iterative steps of the human-centered design process. These steps align with the National Institutes of Health Stage Model for behavioral intervention development (53). The National Institutes of Health stage corresponding to each step in the human-centered design process is indicated in parentheses:

Figure 1.

Human-centered design process. Copyright ©2018, M. Weinger, R. Beebe, and Vanderbilt University Medical Center. All rights reserved.

1. Understand the environment of use (Stage 0).

2. Define user needs (Stage 0).

3. Co-design solutions with users (Stages I and II).

4. Evaluate the designs (Stages II, III, and IV).

5. Review the system after implementation (Stages IV and V).

Although human factors engineering, human-centered design, and patient-centered research are related concepts, there are important distinctions between these terms. Human factors engineering is a scientific and practical discipline focused on understanding human interactions within sociotechnical systems to optimize human well-being and system performance (50). For instance, human factors engineers study how people make decisions in crisis events (54) and how teams collaborate across care settings (55). Human-centered design is a human factors engineering methodology used to develop tools, processes, and technologies that better support humans’ needs within a system through iterative cycles of design and formal evaluation. Human-centered design has been applied across many applications and to a broad range of roles (eg, oil rig workers, pilots, physicians, patients). Human-centered design is recommended by the International Patient Decision Aids Standards collaboration for the development of patient decision aids (56). In contrast, patient-centered research can be any type of research (using qualitative or quantitative methods) focused on promoting outcomes that matter to patients, such as including quality of life measurements during clinical trials (57).

Benefits of human factors engineering

Human factors engineering has been applied to improve safety and quality in many complex systems, such as nuclear power, aviation, and health care, including in the design of medical devices (58,59) and decision support tools (48,60). For instance, Russ et al. (61) applied human factors engineering methods and principles to redesign a decision support tool for medication alerts. In a scenario-based simulation, they demonstrated that the human factors engineering–based tool improved clinician efficiency (ie, reduced time to address alert) and reduced prescribing errors compared with the current tool. Human factors engineering has also been applied to improve code blue medical emergency management (62) and intravenous drug administration in the operating room (63) and to reduce disparities in onboarding for a patient portal (64). In contrast, a lack of human factors engineering in design has led to disaster (65-67). The primary drawback of including a human factors approach is the additional time (and therefore cost) required up front to adequately understand the environment of use and user needs. There is growing evidence that the long-term benefits of human factors engineering to both clinician efficiency and patient safety are worth this initial investment (68).

How to begin to address the challenges of shared decision making in breast cancer: a human factors–based case study

To further illustrate the complexity of integrating decision aids into clinical workflows, we describe an example of a hypothetical patient with breast cancer making a surgical treatment decision. This case is typical of the events and workflow that patients with breast cancer experience (69). Betty is a 55-year-old woman with no family history of breast cancer. Her annual mammogram identified an abnormal mass that was subsequently biopsied; the results confirmed invasive ductal carcinoma. The pathology report was sent to the oncologist through the electronic health record system and to the patient through the patient portal. Betty is scheduled to meet with both a surgical oncologist and a medical oncologist in 3 days in the breast center to review the biopsy results and discuss next steps. At this appointment, Betty and her husband meet with the surgical oncologist, who walks through the pathology report findings (eg, the patient has invasive, estrogen receptor–positive/progesterone receptor–positive, HER2-negative breast cancer of low grade and low proliferative rate). The surgical oncologist also performs a physical exam to determine the tumor’s size and location and to search for lymph node involvement. Betty has already been scheduled for a magnetic resonance imaging (MRI) scan that day after her oncology visit to obtain more detailed information about the mass. Pending the MRI results, the surgical oncologist explains that Betty will likely have the option of either a lumpectomy or a mastectomy. The surgical oncologist explains that these 2 options are equivalent in terms of recurrence risk and overall survival as long as the lumpectomy is followed by radiation therapy. Betty, her husband, and the oncologist all want to get the surgery scheduled today to prevent any treatment delays. Neither Betty nor her husband completely understands what radiation therapy is, but Betty thinks that she would prefer a lumpectomy because the recovery time is shorter. They schedule the lumpectomy, and the surgical oncologist says she will call the patient with the results of the MRI. The surgical oncologist specifically states that the treatment plan may need to be modified depending on the findings. Betty and her husband then proceed to the appointment with a medical oncologist to discuss adjuvant therapy options. Betty gets her MRI before leaving the breast center.

Although this workflow may seem straightforward, there are several opportunities for breakdowns and issues that would hinder the use of a decision aid to foster better shared decision making (Figure 2). For instance, several studies have found that surgeons prefer to have patients view decision aids before the visit, but this timing may not be optimal from the patient’s perspective because the patient may have not yet been told about the cancer or discussed the diagnosis with a physician. In contrast, patients could receive the decision aid after meeting with the surgeon and discussing the pathology report. This point, however, is often too late because the patient may have already made a treatment decision and scheduled the surgery during the visit. If the decision aid is used during the patient’s conversation with the surgeon, the tool must support both patient and clinician workflows and informational needs while not adding time to the busy visit.

Figure 2.

Example of breast cancer surgical treatment decision making in the context of a typical workflow that also shows potential complexities of decision aid integration that hinder rather than enhance patient-centered cancer care.

Human factors engineers can help identify and address these issues by

• modeling and understanding existing complex workflows, including the constraints of organizational policies and procedures (eg, number of visits to discuss decision making with patients), tools and technologies available (eg, the patient portal or secure messaging), and the teamwork involved, including communication and coordination in the process (steps 1 and 2 in Figure 1);

• collaborating with patients, their family members, and clinicians to collaboratively design decision aids and develop optimal workflows to support decision making (step 3 in Figure 1);

• testing and evaluating decision aids before their implementation to ensure that they are sufficiently usable and acceptable to users (step 4 in Figure 1); and

• evaluating workflow issues that arise from the implementation of the new decision aid and identifying design modifications to improve the tool’s fit in the complex sociotechnical (work) system (step 5 in Figure 1).

In addition to improving the design of decision aids, human factors engineers can help redesign workflows to support shared decision making more broadly. For instance, the case example (Figure 2) describes a patient making treatment decisions in 1 visit, but this workflow may be suboptimal because it creates a false sense of urgency for patients to make a decision that same day. Human factors engineering methods should be used to study and redesign processes of care to enhance the overall shared decision-making process. This approach has been applied to improve other complex health-care processes such as family-centered rounds for pediatric patients in the hospital (70).

Although we have focused on workflow-related barriers, it is worth noting that numerous other factors can hinder the use of decision aids, including a lack of financial incentive, limited physician training, and poor infrastructure for decision support systems (11,31). In a broad human factors engineering analysis, these factors would also be investigated and, if possible, optimized. Another challenge is determining who is responsible for maintaining and updating decision aids that have been developed, typically within grant mechanisms. Human factors engineers can also help address these issues. For instance, summative usability testing, a key part of human-centered design, can provide evidence of the benefits of decision support tools to patients (eg, improved satisfaction, reduced decision regret) and to clinicians (eg, improved efficiency), which may motivate health systems to adopt and maintain decision support tools following implementation.

Next steps

Given the ample evidence that decision aids can enhance patient-centered cancer care, future research must focus on how to actually implement effective tools into routine practice (13). Human factors engineering expertise will be vital in this endeavor. Based on the workflow-related issues described here, we propose several next steps to improve decision aid integration and achieve more patient-centered care:

• Include human factors research on team decision making in cancer care to unpack the complex decision-making workflows and needs of decision aid users; this work should include evaluating shared decision-making workflows broadly and identifying how to redesign the process to support patients, their family caregivers, and clinicians.

• Pursue funding for multidisciplinary teams of human factors experts, implementation scientists, clinicians, and informaticians to develop and implement decision support tools.

• Increase involvement of users, including patients, family caregivers, and clinicians, in the design of decision support tools and processes for shared decision making, including consideration of how to support caregiver involvement and diverse clinical roles.

Shared decision making can improve the quality and patient centricity of cancer care, yet numerous workflow challenges limit shared decision making in routine clinical care. Human factors engineering brings a unique perspective that may help shared decision making become a reality. Future research should use human factors engineering expertise in the design and implementation of decision support tools to support complex cancer decisions.

Data availability

No new data were generated or analyzed in this manuscript.

Author contributions

Megan E. Salwei, PhD (Conceptualization; Project administration; Writing—original draft; Writing—review & editing), Jessica S. Ancker, PhD, MPH, FACMI (Conceptualization; Supervision; Writing—review & editing), Matthew B. Weinger, MD, MS (Supervision; Writing—review & editing).

Funding

Research reported in this publication was supported by the Agency for Healthcare Research and Quality through award No. K12HS026395 and No. K01HS029042 awarded to M.E.S.

Conflicts of interest

The authors have no conflicts of interest to report.