We interviewed a wide range of cancer care providers involved in the delivery of targeted cancer therapy in diverse community-based cancer programs, including those not typically included in precision oncology implementation research. Like previous studies of academic and international programs,67, 68 we found similar capability and opportunity constraints in community oncology programs. However, our study extends existing literature by highlighting a larger range of motivational determinants that can facilitate but also slow adoption of targeted therapy, and potentially, other healthcare innovations. Leveraging these determinants may lessen large institutional investments currently considered necessary to meet perceived physician capability deficits.
Community oncology program participants acknowledged knowledge and skill deficits in testing and treatment, especially given the rapid developments in the field, as previously reported.68, 69 Like academic oncologists, community oncologists in this study were comfortable communicating the rationale for targeted therapy to their patients,69 and some even described creative analogies for doing so. Awareness of targeted testing and treatment among our sample appeared high, but participants were less confident in their “how-to knowledge,” or their ability to apply appropriate knowledge about testing and treatment options in practice.69 However, rather than advocate for more education to fill knowledge gaps, an implementation strategy to which technology developers often default, participants in this study suggested institutional-level standardization of testing aligned with clinical practice guidelines and results reporting and treatment education which prioritizes actionable mutations to overcome capability barriers. Others have characterized the actionability gaps in precision medicine68 and called for research to enhance clinical utility.70 Our findings suggest that treatment decision makers prefer prioritization of actionable mutations in results reporting, consistent with existing guidelines for genomic test reporting71, 72 which call for grading the clinical significance of results and limiting reports to the histological and clinical context.71 Opportunities to improve result communication, consistent with previous research73–75 remain.
In addition, our findings further extend understanding of capability determinants in that few sites reported monitoring their testing and targeted therapy use, making it unclear whether their efforts were successful or equitable. Most practices did not have the necessary measurement tools, staffing, or infrastructure to monitor their own performance and thus may not have the performance knowledge needed to regulate their behavior. A limited number of measures related to genomic or molecular testing and treatment are available, required by accreditation agencies, and routinely included in cancer registries,76 77 thus support to develop and implement such monitoring at an institutional level may be needed. Sites used known strategies for improving individual knowledge and treatment decision support but lacked inter-team processes to standardize testing across the eligible patient population.
Similar to prior research, our findings highlight significant opportunity barriers to targeted therapy use, namely the high cost of both testing and treatment, that have long been perceived as barriers to use.41, 78, 79 However, our findings also reflect recent transitions in reimbursement which decrease patients’ out-of-pocket expenses for testing,80 81 39 and assign responsibility for billing to pathology laboratories, shifting incentives for testing from physicians to hospital cost centers39 and creating new organizational landscapes. In addition, we found that most community cancer programs have made organizational and personnel changes to ensure delivery of costly targeted therapy to all eligible by repurposing highly skilled oncology nurses and pharmacists to manage complicated and time-consuming payor and industry requirements. Other cancer programs designed new organizational units to efficiently manage genomic test procurement, tracking, and reporting. Unlike organizational changes to testing management, whose efficiencies may benefit the organization, the addition of new reimbursement and treatment acquisition roles required to deliver targeted therapy to patients unable to cover the costs, sit outside of existing reimbursement structures. Thus, healthcare organizations bear the cost of these activities with little direct benefit to the organization, except perhaps reputational prestige. Nevertheless, the fixed cost of these new activities no matter how streamlined can only be borne by practices with high volume and specialized service, potentially creating disparities among smaller institutions and community practice, where oncologist consider themselves to be generalists.
Across the sample there was steadfast intention and high motivation to provide targeted therapy to cancer patients eligible for it, as was recently reported.68 Nonetheless, our study documents differences in other motivational domains that may be important, namely concerns about the high cost-benefit ratio of testing and treatment, and role identification of the professionals involved in it. Testing is perceived to have societal benefit, allowing for stewardship of costly treatments, in addition to patient benefit. In contrast, the benefits of targeted therapies are not as universally regarded. Although they vastly improve treatment for the few patients eligible for them, they do not cure disease and costs per dose and per course are perceived to be high for both patients and for the institutions delivering them. These beliefs could ultimately influence perceptions of who should bear the cost of organizing coverage. Currently, the treatment stipends are seen as generous and as offering the uninsured wider access to targeted therapy than insured patients. However, because pharmaceutical companies realize the benefit of these programs by increasing their market share, they could potentially balance the lack of societal benefit by working across companies to standardize copayment programs, make eligibility criteria explicit, and broaden qualification.
To our knowledge this study is the first to illuminate ambiguity about who should initiate genomic and molecular testing for cancer. Our study suggests that targeted therapy delivery is difficult because it requires incorporating the new task of genomic test ordering and interpretation into the work scope of professionals typically responsible for treatment decision-making, delivery, and monitoring.82 Most cancer programs relied on oncologists to order somatic tests, the purpose of which is to fully stage the tumor to ensure treatment is appropriate for the patient. However, because the oncologists’ role is focused on treatment, the role of staging the tumor may be at odds with their typical responsibility. Whereas for pathologists, who see themselves as “stewards of the tissue,” definitively staging a tumor falls within existing responsibilities.81 It also aligns with their need to allocate scarce tissue optimally, making pathologist-centered implementation strategies very promising. Some sites had instituted pathologist-initiated test ordering for guideline-recommended tests. So called reflex testing, or automatically ordering one or more secondary tests based on preset criteria applied to the initial test, has been demonstrated to have numerous benefits, including: increasing testing rates83 and identification of mutations or other molecular abnormalities;84, 85 reducing unnecessary testing,86, 87 unnecessary care,88 disparities in care;89 and time to treatment;83, 90 and improving outcomes91 and healthcare operations.92 It has been shown to be cost effective93 and to reduce costs,94 95 mainly by focusing on testing for approved and clinically actionable molecular alterations.
Our description of process from multiple team members’ perspectives, specification of testing and treatment as two distinct behaviors, and comprehensive elicitation of all motivational domains adds new understanding of the strong facilitators and unique barriers cancer care providers experience. In particular, our identification of how determinants of testing behavior differ from the determinants of treatment behavior is a unique contribution not only to understanding precision medicine implementation, but also to the field of Implementation Science. By contrasting the determinants of testing with those of treatment, we uncovered unique patterns of determinants and opportunities for intervention to respond to areas of significant delay in dissemination and implementation. Others have distinguished testing as a process outcome separate from precision medicine application.67 However, specifying testing and treatment as two separate behaviors, each with their own determinants, allows us to consider the different teams involved and connect efforts currently siloed in the fields of pathology and oncology. Although careful specification of implementation strategies is widely encouraged across the field,96, 97 96, 98 less emphasis has been placed on careful specification of intervention behaviors in assessing the behavioral determinants which the implementation strategies are designed to overcome. Instead, most frameworks emphasize understanding the contexts in which innovations are implemented. For example, the Consolidated Framework for Implementation Research (CFIR) emphasizes adapting for context as key to successful implementation, rather than thorough specification of the behavior to be changed.99 CFIR is not unique; the relevance of context is pervasive throughout implementation determinant frameworks.100 In a survey validation of domains identified in the TDF, the Nature of the Behavior construct was dropped from the Framework as it aligned statistically as a separate task, apart from other determinants.49 Although Cane, O’Connor and Michie adamantly emphasized that understanding the nature of behaviors is key to analyzing implementation and other behavior change, they removed the construct from the TDF. Instead they included behavior specification as one of the 8 steps in intervention design in their complementary Behavior Change Wheel (BCW) approach.49 Influenced by the BCW in a previous study in which we identified a promising implementation strategy,101, 102 we subsequently have used careful specification of complex cancer care delivery behaviors to uncover previously unreported determinants,103 including in this study. Our use of process mapping, particularly the Rummler-Brache Diagram or Swim Lane Diagram,53, 54 as a tool to specify the behavior may be a unique contribution, but should be tested as a potentially fruitful addition to determinant analysis.
Limitations. Our study was designed to elicit barriers and best practices related to somatic alterations in tumor tissue. It was not intended to elicit barriers to genetic testing for inherited risk. Hereditary testing typically informs a patients’ prognosis, or risk of disease, rather than being predictive, whereas somatic alterations arising in the tumor can determine whether a treatment will be effective or not. Although some hereditary testing has received FDA approval for treatment decisions, we did not focus on heredity testing. We understood physicians to perceive these two types of tests to have different utility. But because there remains confusion between prognostic and predictive testing and blurring in FDA-approved uses, additional research to assess understanding of, and concerns about, these two types of tests are warranted. Secondly, although we presented our findings to community oncologists and shaped our interpretation by their reactions, we did not formally conduct member checking104 to ensure credibility of results with participants in this study. Instead, we conducted a subsequent survey among a different sample of pathologists, reported elsewhere, that echo many of the concerns about testing among pathologists we report here. Likewise, our study was designed to comprehensively elicit a broad range of barriers and best practices. Salience of each construct was not evaluated. Future studies using representative sampling could narrow these constructs to those deemed most important to the majority of community oncologists. Finally, the high motivation and intention to use targeted therapy among provider participants in this study may reflect sampling bias introduced by our initial effort to require data sharing. However, a recent study68 confirms our finding regarding high motivation to use targeted therapy in the current era and these differences in motivation between our study and earlier studies may reflect changes in trends over time, rather than groups of oncologists who hold discordant views.