INTRODUCTION

Precision medicine considers individual differences and preferences to inform personalized health-care decisions that aim to improve health outcomes. Genomic medicine is fundamental to current precision medicine programs. Yet, many health-care systems have not systematically integrated genomic medicine into clinical care because of the complexity of genomic information, limited numbers of genetics professionals, lack of preparedness of most health-care professionals, and barriers to coverage and reimbursement.1,2,3,4,5,6,7,8 An optimal implementation plan in genomic medicine is crucial to realizing the benefits of precision medicine services delivered within health-care organizations. Other applications of precision medicine, such as proteomics and metabolomics, will probably face similar challenges to adoption.

Like other health-care systems, the Veterans Health Administration (VHA) is considering how best to integrate research discoveries into clinical practice, and expand clinical genetic and genomic medicine services.9 The VHA has made a substantial investment in precision medicine, as evidenced by the establishment of the Million Veteran Program (a research program designed to study how genes affect health) and formation of the national Genomic Medicine Service (a centralized team of genetic counselors based in Salt Lake City who provide genetic services through telehealth service agreements). In addition, there are regional VHA clinical genetics programs based in Boston, Houston, and Los Angeles that provide genetic consultation on-site and via telehealth. To assist the VHA with precision medicine strategic planning, we convened key stakeholders for a one-day conference. The goal of the conference was to foster partnerships between key stakeholders for agenda setting to inform precision medicine policy, delivery, and research. We used the conference proceedings as a source of qualitative data to gather themes for the development of a logic model that could be used to inform precision medicine implementation in the VHA and other health-care organizations.

MATERIALS AND METHODS

Conference planning and participants

We invited participants from stakeholder groups identified as being critical to achieve the conference goal, including: patients, family members, and advocacy organizations; clinicians (primary care providers, geneticists, genetic counselors, and other medical specialists); researchers (basic science, health services, and implementation researchers, including information technology experts); policymakers and payers; and administrators and managers of VHA and non-VHA health-care organizations. We used purposive recruitment to achieve a diverse participant panel that would provide a comprehensive view of precision medicine within and outside the VHA. There were 90 invitees, about half of whom were affiliated with the VHA.

The conference was held on 25 August 2016 in Arlington, Virginia. Each of nine sessions comprised a panel of three to six speakers and time for an open discussion with the audience. Before the conference, session moderators emailed speakers three to four open-ended questions to inform their presentations. These questions addressed the benefits of (and opportunities for) precision medicine, challenges and barriers to precision medicine, and strategies that can promote precision medicine adoption and implementation. We used qualitative analysis of the conference proceedings to inform the development of a logic model, and applied implementation science frameworks to provide context and anchor themes that informed the model, as described below. The project was considered non-research by the Veterans Affairs Greater Los Angeles Healthcare System Institutional Review Board.

Research topics identified

At the conference, during the afternoon break, three program committee members (M.T.S., B.L., and J.P.) met with one of the invited speakers (E.M.Y.) to identify research topics addressed during the conference. This group identified nine research topics relevant to the planned thematic analysis: preclinical and clinical research; health services and outcomes research; dissemination and implementation research; organization and provider behavior research; partnered research; economics research; information technology and clinical informatics; public engagement; and ethics and equity.

Data collection and transcript coding

The conference proceedings were audio-recorded and professionally transcribed. Transcripts were reviewed and edited for accuracy. Thematic analysis of the transcripts was based on an iterative, consensus approach that included: establishment of a coding dictionary; independent dual coding of themes with comparison (C.C.-C. and M.T.S.); deliberation and resolution of discrepancies; and review by an auditor not involved in coding when resolution of coding discrepancies could not be reached (J.P.).

We used a matrix method to facilitate transcript coding. We listed speaker statements in the rows of the matrix and the nine research topics in the columns of the matrix. The coders matched each statement to a research topic and entered a theme into the intersecting cell of the matrix. When statements mapped to more than one research topic, the coders selected primary and secondary research topics. The primary research topic was used in the synthesis of findings. We also characterized each statement and associated theme as an opportunity, challenge, or strategy. We defined opportunities as existing activities and resources that could support precision medicine implementation, and challenges as any factor, objective, or subjective perceived by the stakeholders as a potential barrier to precision medicine implementation. For implementation strategies, we used the definition from Proctor and colleagues, “methods or techniques used to enhance the adoption, implementation and sustainability of a clinical program.”10 We excluded statements of casual conversations (e.g., transition between speakers), and those related to the speakers’ personal information or role in precision medicine.

Anchoring precision medicine themes to implementation science frameworks

Figure 1 illustrates the process we used for coding and mapping that resulted in a precision medicine implementation plan. Two coders (C.C.-C. and M.T.S.) independently mapped the themes identified to the W.K. Kellogg Foundation Logic Model components,11 Consolidated Framework for Implementation Research (CFIR) constructs,12 and Expert Recommendations for Implementing Change (ERIC) compilation of implementation strategies.13 Discrepancies were discussed and, when needed, adjudicated by a third investigator (J.P.).

Fig. 1: Flow chart summarizing the coding and analysis of the conference proceedings.
figure 1

CFIR Consolidated Framework for Implementation Research, ERIC Expert Recommendations for Implementing Change.

Full size image

The logic model consists of the planned work (inputs and activities) and intended results (outputs, outcomes, and impacts) for implementation of a program. We used the logic model component definitions, as described by the W.K. Kellogg Foundation. Inputs include the human, financial, organizational, and community resources available to do the work; activities include the processes and intentional actions to bring about the intended program results; outputs are the direct products of program activities; outcomes are changes in behavior, knowledge, skills, status, and level of functioning; and impact is the intended or unintended change occurring in organizations, communities, or systems as a result of the program activities.11

We mapped themes stated as opportunities or challenges to the CFIR, as defined by Damschroder and colleagues.12 The CFIR provides a menu of constructs that can be used as a practical guide for systematically assessing potential barriers and facilitators in preparation for implementing an innovation and providing theory-based constructs for developing context-specific logic models. The CFIR is composed of five major domains and its associated constructs: 8 constructs comprise the intervention characteristics domain (e.g., evidence strength and quality), 4 constructs comprise the outer setting domain (e.g., patient needs and resources), 12 constructs comprise the inner setting domain (e.g., culture and leadership engagement), 5 constructs comprise the individual characteristics domain, and 8 constructs comprise the process domain (e.g., plan, evaluate, and reflect). We chose the CFIR because it is easily customized to diverse settings and scenarios. By providing a framework of constructs, the CFIR promotes consistent use of constructs, systematic analysis, and organization of findings from implementation studies.12 All CFIR domains are relevant to precision medicine implementation; however, for this study, the inner domain constructs are of particular interest given the focus on implementation within a health-care organization.

We mapped themes stated as strategies to the compilation of implementation strategies, as defined by the ERIC study.13,14,15 The ERIC compilation consists of 9 clusters of 73 implementation strategies, including: use evaluative and iterative strategies (e.g., audit and provide feedback, and develop and organize quality monitoring systems); provide interactive assistance (e.g., provide local technical assistance, and provide centralized technical assistance); adapt and tailor to context (e.g., use data experts and use data warehousing techniques); develop stakeholder inter-relationships (e.g., use advisory boards and work groups, and identify early adopters); train and educate stakeholders (e.g., develop educational materials and conduct educational meetings); support clinicians (e.g., remind clinicians and facilitate relay of clinical data to providers); engage consumers (e.g., involve patients/consumers and family members); utilize financial strategies (e.g., access new funding and alter incentive/allowance structures); and change infrastructure (e.g., change record systems and change service sites). We chose the ERIC compilation because it is a comprehensive collection of implementation strategies developed by implementation science experts using a systematic approach to identify, develop, and evaluate a consistent language and description of implementation strategies.13

RESULTS

Eighty individuals attended the conference and 58 contributed to the discussion as speakers or audience members. We identified 419 discrete speaker statements. We excluded 116 statements describing participant background and experience with precision medicine. Among the remaining 303 statements, we identified 47 unique themes. We classified 82 (27%) statements as opportunities, 89 (29%) as challenges, and 132 (44%) as strategies.

A logic model for precision medicine implementation

A logic model for a precision medicine program within a health-care organization is shown in Figure 2. Assumptions underlying the logic model discussed during the conference included: precision medicine will improve health outcomes; precision medicine has value beyond clinical utility; and the confluence of clinical care and research is critical for successful precision medicine implementation. As we mapped the conference themes to the logic model components, higher-level concepts emerged, and we arranged the themes within these concepts. For example, three concepts emerged for inputs: infrastructure, big data, and resources.

Fig. 2: Logic model for precision medicine implementation informed by key stakeholders.
figure 2

Concepts and associated themes identified from transcript coding of the conference proceedings were arranged into a logic model for precision medicine implementation. The logic model can be used by a health-care organization to inform the planned work (inputs and activities) and intended results (outputs, outcomes, and impact). Assumptions underlying the logic model include: precision medicine will improve health outcomes; precision medicine has value beyond clinical utility; and the confluence of clinical care and research is critical for successful precision medicine implementation. IT information technology, PM precision medicine, CER comparative effectiveness research, ELSI ethical, legal and social issues, HCS health-care system, ROI return on investment.

Full size image

The contribution of the different stakeholder groups to the logic model concepts is shown in Figure 3. Individuals from every stakeholder group participated in the discussion. Individuals from multiple stakeholder groups contributed to the themes that mapped to every concept within the logic model, except for the health-care utilization concept within the outcomes component. In this instance, only clinicians contributed statements.

Fig. 3: Number of participants contributing to the logic model concepts by stakeholder group.
figure 3

Total number of participants by stakeholder group: administrators/managers: n = 11; clinicians: n = 22; patients/family/advocates: n = 6; policymakers/payers: n = 5; researchers: n = 13. One participant could not be identified in the transcript, and was therefore not featured here.

Full size image

Research topic mapping to the logic model

Many of the conference discussions addressed the early stages of the logic model, with 73% of the themes mapping to inputs (e.g., big data) and activities (e.g., precision medicine practice and precision medicine research). The research topics of information technology and clinical informatics, outcomes research, and organization and provider behavior research predominated within the inputs component of the logic model, with most themes stated as opportunities. For the activities component, partnered research, dissemination and implementation research, and organization and provider behavior research were the most common, with partnered research most often stated as an opportunity, and dissemination and implementation research and organization and provider behavior research most often stated as a challenge or strategy. The outcomes research topic predominated among statements mapping to the logic model components of outputs (most often stated as opportunities) and outcomes (most often stated as challenges). For the impact component of the logic model, the research topics of dissemination and implementation research, outcomes research, and ethics and equity were most common, and typically stated as challenges. Supplementary Table 1 shows the distribution of research topics for themes stated as opportunities, challenges, and strategies according to the logic model components and concepts.

Context for precision medicine planning: mapping CFIR constructs to the logic model

We mapped themes stated as opportunities or challenges to 24 CFIR constructs (Table 1). The logic model concept of precision medicine research, which had the greatest number of themes stated as opportunities for precision medicine (e.g., communication with stakeholders, partnerships within a health-care organization, and partnerships across sectors) mapped to the CFIR constructs of: evidence strength and quality, adaptability, cosmopolitanism (i.e., the degree to which an organization is networked with other external organizations), networks and communications, implementation climate, available resources, and planning. The logic model concept of equity and access had the greatest number of challenges for precision medicine (e.g., ethical, legal and social issues that limit precision medicine practice, such as return of genetic test results in the research setting, and individuals with rare genetic disorders being a vulnerable population) and mapped to the CFIR constructs of complexity, patient needs and resources, external policy and incentives, relative priority, access to knowledge and information, and knowledge and beliefs about the intervention.

Table 1 Conference themes stated as opportunities or challenges mapped to CFIR and the logic model
Full size table

Context for precision medicine planning: mapping ERIC implementation strategies to the logic model

We mapped themes representing strategies for precision medicine to 42 implementation strategies from all 9 clusters of the ERIC compilation (Table 2). The logic model concept of precision medicine practice had the greatest number of themes stated as strategies, whereas health-care utilization had none. Many strategies were related to information technology and clinical informatics to support clinical decision-making and reduce variation in precision medicine practice, such as changing record systems, using data warehousing techniques, developing and implementing tools for quality monitoring, and using centralized technical assistance. Other strategies focused on promoting communication and data sharing by developing partnerships, engaging consumers, and training and educating stakeholders. To improve access to precision medicine, conference participants described the use of financial strategies, changing service sites, creating new clinical teams, revising clinical roles, recruiting and training leadership, adapting and tailoring precision medicine interventions to context, and mandating policy changes.

Table 2 Conference themes stated as strategies and mapped to the ERIC compilation of implementation strategies and logic model
Full size table

DISCUSSION

We have developed a comprehensive logic model for precision medicine program planning that is grounded in implementation science and informed by deliberations of diverse stakeholder groups brought together for a conference focused on integration of precision medicine in clinical care. The use of the W.K. Kellogg Logic Model helped organize the themes and higher-level concepts described by conference participants. The logic model serves as a roadmap for precision medicine program planning and evaluation that links short- and long-term outcomes with inputs and activities.11 We also used implementation science frameworks (the CFIR and ERIC compilation) to provide context for the logic model.12,13 This formalized approach provides a common set of constructs that are transferrable to any health-care organization considering precision medicine implementation, and is useful for standardizing investigations of precision medicine implementation studies. The mapping of CFIR domains and constructs to the precision medicine challenges and opportunities could be used as a checklist to help a health-care organization identify local factors likely to influence precision medicine implementation, and mapping of the ERIC implementation strategies to the precision medicine strategies could help with prioritizing strategies likely to enhance the adoption and implementation of precision medicine across the organization.

Planned work for precision medicine implementation (i.e., logic model inputs and activities) predominated the conference discussion, which suggests we are still early in the adoption of precision medicine. We learned that while integrated health-care systems and information technology and clinical informatics are inputs that provide opportunity for precision medicine implementation, precision medicine practice faces fundamental challenges that must be overcome to realize the promise of precision medicine. These challenges include an unprepared health-care workforce, the need for evidence-based clinical guidance, and the need for novel care models that address the diversity of the US population, and provide equitable access, improve care coordination, and reduce variation in precision medicine practice. Additionally, while there is considerable enthusiasm about the promise of precision medicine outputs (e.g., diagnosis, risk assessment, and targeted treatment) and cautious optimism about outcomes from precision medicine interventions (i.e., clinical utility, personal utility, and targeted health-care utilization), the longer-term impacts of precision medicine are fraught with challenges relating to uncertainties about the value of precision medicine, equity of and access to precision medicine, cost-effectiveness, and return on investment.

We used the CFIR to systematically assess potential barriers and facilitators to precision medicine implementation, which can facilitate collaborations and evaluation across organizations.12 The Implementing GeNomics In PracTicE (IGNITE) Network’s Common Measures Working Group also used the CFIR to develop standardized measures to inform implementation research.16 However, the IGNITE work group found that the CFIR did not capture factors relevant to community values for patients and families, such as “understanding patient perceptions, anxiety, and personal utility.” We also identified themes relevant to patient experience and personal utility. However, our perspective was that of a health-care organization and we viewed patients as stakeholders within the health-care organization. As such, we mapped patient experience within the CFIR domain, "characteristics of individuals", and personal utility within the CFIR constructs of "evidence strength and quality", and "relative advantage" within the domain, "intervention characteristics".

We identified 42 discrete implementation strategies from the ERIC compilation that leverage the opportunities and address the challenges for precision medicine implementation. All conference themes representing a strategy were depicted within the ERIC compilation of implementation strategies. However, the saliency of the strategies described for precision medicine implementation may differ from other innovations. As described by Waltz and colleagues, a panel of implementation science and clinical experts rated the importance and feasibility of all implementation strategies included in the ERIC compilation.15 The goal was to help facilitate the selection of strategies that are best suited for implementation efforts. These experts gave low ratings for both importance and feasibility to the strategies within the cluster of "change infrastructure" (e.g., change record systems, change physical structure and equipment, and change service sites). These strategies frequently came up as key during the discussions regarding precision medicine implementation. Conference participants identified information technology and clinical informatics infrastructure changes and integration of data analytics as critical to successful precision medicine research, effective precision medicine practice, and facilitating a learning health-care system. They identified changes in service sites, such as embedding precision medicine research within clinical precision medicine practice, as necessary to promote new care models that could increase access to and decrease variation in precision medicine practice. Thus, while these strategies may have low feasibility, they appear to have high importance to successful precision medicine implementation.

Many of the precision medicine implementation strategies we identified in the analysis of the conference proceedings have been described by others.2,5,6,7,8 However, the contribution of diverse stakeholders facilitated the development of a wide range of precision medicine implementation strategies, some of which have not been reported previously, such as using centralized technical assistance for genetic consultation through the use of telehealth to reduce variation in precision medicine practice, and using formularies for genetic tests to improve access to precision medicine by facilitating coverage and reimbursement. Moreover, to our knowledge, this is the first attempt to characterize a comprehensive list of precision medicine implementation strategies according to the ERIC compilation to facilitate precision medicine program planning and evaluation.

The IGNITE network has published work describing seven ERIC implementation strategies addressing challenges that are common to the six implementation projects within their network, most of which we describe (i.e., use data warehousing techniques, conduct educational meetings, prepare patients to be active participants, and involve patients) and some we have not (i.e., develop educational materials, conduct educational outreach in clinical settings, and use mass media to engage patients).8 While these implementation strategies utilized by IGNITE were not explicitly mentioned by the conference participants, several other strategies regarding provider education and patient engagement were mentioned. Thus, our wide-ranging list of implementation strategies that includes strategies from all clusters within the ERIC compilation is not exhaustive, but is satisfactory for contextualizing the logic model components and concepts that can be used for precision medicine program planning and evaluation.

Research topics mentioned during the conference provide the basis for a research agenda that could fill the evidence gaps for precision medicine. The research areas we identified span the continuum of translation research for genomic medicine, as described by Khoury et al.17 Yet, the need for health services, outcomes, and dissemination and implementation research dominated the conference discussions, consistent with the conference goal, and in keeping with previous work showing limited funded research and publications in these areas.18

We learned that preclinical studies are still needed to define genotype–phenotype associations. For success, data sharing is critical and could be achieved with partnered research though collaborations within health-care organizations and across sectors, and through public engagement to promote interest in precision medicine. Health services and outcomes research are needed to generate robust evidence for personal and clinical utility resulting from precision medicine interventions, and to better understand the contextual factors that influence the usefulness of precision medicine interventions. Dissemination and implementation, organizational, and provider behavior research are needed to help improve access to precision medicine and reduce variation in precision medicine practice. It will be essential to develop data standards and address the data extraction needs necessary for a learning health-care system.19 Research is also needed to determine how best to communicate the complexities of precision medicine information and to educate the workforce and public. Lastly, health services and economics research are needed to understand precision medicine care coordination and optimal models for the delivery of precision medicine interventions, barriers to coverage and reimbursement, and cost-effectiveness of precision medicine.

Important limitations to this work need mentioning. We used a structured process with instructions to the conference speakers that constrained their presentations, and while we engaged numerous individuals from a variety of stakeholder groups who contributed to the discussion, other viewpoints may not be represented. We used a predefined set of research topics to assist in thematic coding and we used coders who are knowledgeable about precision medicine research and practice, which may have influenced the thematic analysis. However, themes were derived from the data using dual independent coding to mitigate biases, and we have provided coded paraphrased statements in the tables to ensure transparency.20

Conclusion

We have developed a logic model for precision medicine implementation within a health-care organization informed by key stakeholders that provides a roadmap for precision medicine program planning. Furthermore, we have contextualized the logic model through the use of implementation science frameworks. The domains and constructs from these frameworks provide a basis for prioritizing implementation strategies that can leverage facilitators and address barriers to enhance the adoption of precision medicine within a health-care organization. This formalized approach helps to organize and standardize precision medicine implementation and evaluation. Integrated health-care systems offer opportunities to realize the promise of precision medicine through the use of population-based data that can generate evidence regarding precision medicine outcomes and inform clinical decisions at the point of care. Preparing the health-care workforce and educating the public remain paramount to achieving successful precision medicine implementation. Implementation planning should consider novel care models, modes of delivery, and financial strategies to ensure access and reduce variability to precision medicine interventions. Precision medicine programs should reflect the diversity of the populations served and incorporate stakeholder perspectives on the value of precision medicine interventions when evaluating the effectiveness of precision medicine implementation.