This multi-method qualitative study was embedded within a pragmatic trial that tested the implementation of a multicomponent evidence-based intervention aimed at enhancing care coordination for breast and colorectal cancer survivors with chronic conditions [15]. Study procedures were approved by the University of Texas Southwestern Institutional Review Board (STU 102015–090), the University of Texas Health Science Center at Houston, and by Parkland Health Office of Research Administration, and reporting follows the Standards for Reporting Qualitative Research guidelines [16].

Setting

This study was conducted at Parkland Health (Parkland), the safety-net health system serving Dallas County, TX, USA [17, 18]. “Safety-net” healthcare systems are those that deliver healthcare primarily to uninsured, Medicaid, and other low-income and vulnerable patient populations [19]. Parkland includes a network of 13 primary care clinics located in predominantly under resourced, ethnic/racial minority communities across Dallas County, and a centrally located main campus. The main campus consists of an inpatient hospital, outpatient surgery center, and specialty care clinics, which include multidisciplinary cancer clinics (i.e., medical, surgical, and radiation oncology clinics).

Breast and colorectal cancer are the top two types of cancers treated at Parkland. Twenty-four percent of patients with breast cancer present with stages 3 or 4 breast cancer compared to 10% nationally; 61% of patients with colorectal cancer present with stages 3 and 4 cancer compared to 45% nationally [20].

Evidence-based intervention components and implementation strategies

Project CONNECT was a multicomponent evidence-based intervention and included (1) an electronic medical record (EMR)-based patient registry and (2) a care coordinator [15]. The registry identified patients diagnosed with stages I–III breast or colorectal cancers plus one or more of the following chronic conditions: diabetes, hypertension, heart disease, chronic kidney disease, and/or chronic lung disease. The care coordinator was a registered nurse employed by Parkland who helped connect study eligible cancer survivors to primary care by facilitating appointments with primary care and coordinated care for patients between oncology and primary care. Strategies identified a priori to implement the intervention components into clinical practice including the following: identifying champions, changing records systems, creating new clinical workflows, and flexibility in implementation (Table 1) [15].

Table 1 Intervention components, functions, and implementations strategiesGuiding theoretical and conceptual frameworks

The practice change model (PCM) and the Consolidated Framework for Implementation Research (CFIR) are determinant frameworks that guided data collection to identify barriers and facilitators of implementation [22, 23]. The PCM includes four elements (e.g., internal motivators, external motivators, resources, and opportunities for change) and depicts how these multi-level elements can impact intervention implementation in healthcare settings over time [22]. The CFIR is a menu of individual-, program-, and organizational-level constructs consolidated from 19 theories and models related to intervention implementation [24]. The constructs are organized into five overarching domains: intervention characteristics, outer setting, inner setting, characteristics of individuals, and process. These frameworks are highly complementary. PCM grounded our focus on drivers of practice operations and potential interactions, while CFIR helped us attend to relationships [25] between our intervention, actors in the practice, and the structure and sequence of care delivery to maximize learning from our real-world setting [26].

Proctor and colleagues’ taxonomy of implementation outcomes [27] and the Implementation Research Logic Model (IRLM) [28] informed our data analysis and synthesis [27].This study used qualitative data to assess two implementation outcomes at the patient and provider levels (i.e., intervention acceptability and appropriateness) and two outcomes at the organizational level (e.g., intervention adoption and penetration). Acceptability is defined as patient and/or provider satisfaction with the intervention, and appropriateness is defined as the perceived fit of the intervention in the setting [27]. Adoption is defined as the initial uptake of the intervention; penetration is defined as the integration of an intervention within a clinical team, which is similar to the concept of “reach” in Glasgow’s RE-AIM framework [29]. Adoption and penetration were assessed longitudinally during intervention implementation (Phase 2, see below) allowing assessment of continued adoption or utilization of the intervention beyond initial uptake. Finally, the IRLM is a visualization tool to depict causal pathways between intervention components, determinants (i.e., barriers and facilitators) of implementation, implementation strategies, mechanisms of action, and implementation [28]. Mechanisms of action define how implementation strategies operate to influence outcomes. We used the IRLM to elucidate the relationships between determinants, mechanisms, and implementation outcomes.

Data collection

Trained investigators (R. T. H., P. M. C., S. C. L.) collected qualitative data throughout intervention implementation from September 2016 through June 2020. This included two phases of the study. Phase 1 of data collection occurred from September 2016 to September 2018, pre-intervention implementation. Phase 2 occurred from September 2018 to June 2020 during intervention implementation.

We used purposive sampling to select clinical team members who varied by their roles and specialty to identify barriers and facilitators to delivering coordinated care for patients with cancer and chronic conditions. Study participants included clinicians (e.g., physicians, nurse practitioners), clinic staff (e.g., nurses, care coordinators, social workers, financial services coordinators), and health system leaders (e.g., unit managers, clinic managers, and medical service chiefs) in oncology, primary care, and specialty care. We recruited multiple participants for each role and unit to solicit diverse perspectives.

Data sources included the following: (1) documents, (2) structured observations and field notes, and (3) semi-structured interviews with patients, providers, staff, and leaders from multiple departments across the integrated safety-net system.

Documents

Documents included meeting notes, policies and procedures, correspondence among stakeholders, EMR screenshots, patient-facing materials, tools and checklists, and other resources. Documents were requested from providers, staff, and leaders, and they were also offered unsolicited by interviewees and observed stakeholders to clarify processes, provide supplementary information, or serve as historical records.

Structured observations and field notes

We used structured observation guides to facilitate consistent data capture of care coordination and practice change processes [30]. Exemplar domains and questions included the following: evidence of team-based care (e.g., do oncology providers discuss other conditions or comorbidities?), documentation of practice (e.g., where do oncology providers document information related to the survivorship care plan or referrals for follow-up care after discharge?), patient access to information (e.g., do providers tell patients what to do in the event of acute needs?), continuity of care (e.g., how are subsequent appointments scheduled?), and team-based care (e.g., to what extent do providers engage patients in taking an active role in their care?). We also selected sites for observation to capture the patient pathway and provider/staff movement through the care coordination process (e.g., registration and intake areas, patient-provider interactions, provider-staff interactions and work areas, and nurse navigation, referral, and case management processes) [31].

Semi-structured interviews

Interviews were semi-structured to guide the interviewer through pre-planned topics while allowing for follow-up questions tailored to participant feedback and for additional unplanned questions to be incorporated as appropriate. Interview guides were iteratively developed by investigators and adapted to role and clinical unit. We anticipated barriers and facilitators to implementation based on the CFIR [23] and PCM [22] and, accordingly, focused the interviews on domains including the following: care coordination processes between oncology and primary care, perceptions of the role of the nurse coordinator and registry (interventions), challenges or gaps in care for cancer survivors with chronic conditions, communication about policies and procedures within clinics, EMR documentation and challenges, delineation of roles and expectations between oncology and primary care providers and staff, and patient feedback about areas of confusion or concern. Prior to participating in an interview, informed verbal consent was obtained from all study participants. Patients received a US $25 gift card in appreciation for their time. In accordance with Parkland policy, employees were not provided with an incentive to participate in research.

Data analysisImmersion-crystallization processes

Data analysis proceeded in four immersion-crystallization cycles, or repeated exposure to and synthesizing of data, to identify themes and categories (Supplementary Materials) [32].

In cycle 1, the team developed two deductively driven thematic codebooks based on interview guide topics, pre- and post-intervention phases, and a preliminary review of documents (n = 259 unique documents), field observations (n = 11), and interview transcripts (n = 140). Additional emergent themes were incorporated into the initial codebook drafts for the first 10% of transcripts, and the finalized codebook was used for remaining transcripts. Codebooks for the two phases included many of the same codes; however, each also included additional unique codes given differences in thematic foci and emergent findings during each phase. For example, Phase 1 codes included existing barriers to care coordination, organizational structure, and processes; Phase 2 codes included patient experiences, acuity of care, and transitions in care. All coding was completed in NVivo 12.0 (QSR, Australia). After coding all data, the team created node reports, summaries of data collected, and exemplar quotes for each code and identified codes tying together steps in the cancer care continuum to the intervention components: care coordination, survivorship planning, intervention, and intervention impact.

In analysis cycle 2, the team applied codes from PCM and CFIR to the selected node reports focusing on identifying organizational inner setting characteristics, system resources, stakeholder motivations, and opportunities for change. The purpose of this cycle was to understand how and why care coordination processes occurred pre- and post-intervention. In cycle 3, the team returned to the findings from cycles 1 and 2, coding for in order to describe how the intervention components and care coordination processes mapped to implementation outcomes. Implementation outcomes were assessed from qualitative data; most validated quantitative implementation outcome measures were not available at the start of this study. In cycle 4, the team met weekly to interpret findings and synthesize data linking implementation strategies, determinants, mechanisms, and outcomes using the IRLM.