The DAP supports planned adaptation of evidence-based interventions by distinguishing core components from adaptable features, guiding allowable changes, training, fidelity monitoring, and system adjustments in a planned, rather than ad hoc way. The DAP model involves identifying core elements and adaptable characteristics of an evidence-based practice, then provides specific training on allowable adaptations, processes for fidelity monitoring and support, and identification of solutions to system and organizational adaptations. Thus, the DAP model improves real-world implementation fidelity while supporting iterative change.

Following the Designing for Dissemination and Sustainability (D4DS) framework (5), we augmented the DAP model by engaging core stakeholders and initiating dynamic adaptation processes earlier upstream during the intervention development process, starting with conceptualization and intervention design. By doing so, we effectively extended the DAP from an implementation only study design to a hybrid effectiveness-implementation design. The original DAP model included four sequential phases (6). In our project, we extended DAP by allowing feedback from the early phases of the intervention design to inform subsequent phases of implementation and scale up.

The multi-disciplinary PC-TCM research team included PhD-prepared faculty, doctoral students, and administrative support representing the professions of nursing, social work, physical therapy, implementation science, industrial systems engineering, and computer science engineering. The team was divided into two sections, one supporting the technical aspect of the intervention and the other supporting implementation with the partnering organizations, joined through the HIE intervention. This paper focuses on the work of the implementation resource team (4) which was led by a nurse anthropologist, the principal investigator (PI), co-investigators with expertise in implementation science, systems engineering and social work, a project coordinator, and PhD students in nursing. This team worked intensively with the partners, both technical and clinical, through cross sector workshops, regular monthly meetings, ad hoc meetings over Zoom, and direct observation in the clinical sites. In addition to ethnographic observation and extensive fieldnotes, the team collected qualitative data in the form of periodic reflections from partners (managers and staff) (6) and semi-structured interviews with persons experiencing complex social and medical conditions. Following a participatory design approach, we involved frontline staff and administrators from all participating community sites. This process included engaging community partners early, so they were able to work closely with the implementation resource team across all phases of the design and testing of technological platforms.

To facilitate this process, meetings were held between clinical partners and the teams to closely monitor project milestone achievement and to share insights from ongoing data collection and analysis. The PI coordinated the project, led regular meetings, tracked milestones, and oversaw data sharing processes. Please see Figure 1 for a conceptual overview of the relationships between the research team, community partners, data collection processes, and the HIE.

Throughout the project, we collected and reviewed data from 53 documents (2021–2024), including meeting minutes, transcripts of discussions, and project schedules to evaluate real-time decision-making processes, protocol deviations, near-misses (events potentially leading to implementation failure), and reflective insights on challenges and solutions encountered throughout the project. We also conducted qualitative interviews with research team members and clinical staff and technical staff at partnering organizations, alongside intensive interviews examining the perspectives of people with high medical, social and behavioral needs (6, 9) who were patients of the practice. The interviews were audio recorded; the data were transcribed, categorized, and qualitatively analyzed to identify patterns, themes, and relationships between the implementation process, barriers, and stakeholder roles. The findings of these qualitative interviews, including details regarding interviewee profiles, have been published elsewhere (10, 11).

The project's innovation included both technical and social components. For the technical innovation, we developed an enhanced transitional care alert (ETCA), delivered electronically to primary care and behavioral health providers post-hospital discharge to optimize cross-sector collaboration. The ETCA relied on principles from health informatics, team science, and implementation theory, and HEALTHeLINK was central to the process, automatically pulling aggregated data into the ETCA from the primary care practice (practice roster and high needs report), the discharging acute care facility (discharge notification), and the clinical data repository to support timely outreach from the clinical site. For all patients receiving an admission discharge transfer (ADT) notification who also appeared in the high needs report, HEALTHeLINK aggregated these disparate data sources into a single ETCA. The ETCA information was automatically transmitted to the community-based organizations to facilitate cross-sector collaboration and shared care planning.

The social innovation involved an iterative process of stakeholder engagement and multi-disciplinary teams in the re-design and optimization of practice workflows in response to the ETCA. In this phase, the team worked closely with the clinical staff at the community sites to co-design a revised alert that included information that facilitated a nurse-led outreach phone call within 72 h of discharge. The final ETCA “dashboard” triggered by a hospital discharge, included: (1) contact and demographic information, (2) Adjusted Clinical Groups (ACG) Risk Score, (3) information from the admission, discharge, transfer (ADT) alerts, (4) information about all providers caring for the patient, and (5) relevant chronic/behavioral/social needs information (Figure 2).

We used overlapping qualitative, quantitative, and ethnographic approaches to better understand how to identify and classify individuals with high needs into useful ETCAs for the practice partners. First, we identified relevant codes and cross-walked them to the International Classification of Disease (ICD-10) for chronic conditions following the H-CUP Clinical Classifications Software Refined (CCSR) (12). Then, we developed a clinical algorithm for integrating social determinants of health (SDH) needs into the COMPLEXedex algorithm. The COMPLEXedex is a model for classifying risk based on a hierarchical ranking of 14 chronic health conditions (13–16). The COMPLEXedex-SDH, developed by nurses researchers on our team, integrated social needs identified through qualitative case analysis and classification of social needs was informed by the Gravity Project (17, 18). The COMPLEXedex-SDH identifies individuals with high needs based on having two or more chronic conditions, social risk factors, and care utilization patterns, shifting the focus from costs of care to overall need.

Ethnographic work included formal (workshops and meetings) and informal interactions (observations) with the clinical staff at the practices to inform development of the ETCA. These processes facilitated trust between stakeholders and the research team. By embedding themselves in a sustained, dialectical manner (19), the team helped bridge the persistent gap between academia and real-world implementation (20–22). The project coordinator facilitated communication centrally using email and cross-sector Zoom meetings to enable organizations to safely work out their idea collaboratively, while observing each other in the implementation process. This communication system was essential as both individuals and organizations were observed to be simultaneously espousing negative (this will never work), neutral (just doing the work), and positive (this is going to work great) sentiments, making parallel linear progress difficult. Initially, cooperation occurred without consensus (20), as participants struggled to align perspectives. However, repeated exposure to each other's positions allowed partners to develop shared understanding and collective goals, enabling the intervention to evolve to complexity in real time.

Grounded in equitable implementation and designing for dissemination (5, 28), we prioritized integrating the patient experience [contending with the work of being a patient (“treatment burden”), self-care and self-management, medication management, and interacting with the healthcare system after hospital discharge] (29) into both the intervention and its evaluation. During planning, we interviewed 27 individuals from the practice sites with complex, co-occurring health conditions (e.g., diabetes, behavioral disorders) and unstable housing (10). Participants were recruited and interviewed at a social services organization offering shelter and medical respite care. Later, we interviewed 22 patients who received the nurse outreach calls in the immediate post-discharge period (11).

These conversations revealed that unmet social needs, particularly housing and food, often took priority over medical issues, and these unmet needs heavily influenced treatment burden (34). This information shaped all aspects of the intervention, from the ETCA content to nurse training content to ensure that outreach phone calls addressed social as well as clinical needs. Our findings also informed updates to standard treatment burden measures, ensuring relevance for patients with complex medical needs and social risk factors (11). These perspectives provided critical insights, enriching prior conceptual models, complementing data gathered from other stakeholders, and they shaped the development of the intervention and the implementation evaluation strategies. In complex cross-sector implementation environments, patient perspectives are a key information stream that can inform approaches to preventing failure to rescue the intervention, designing for sustainability, and avoiding implementation drift.

Before the intervention, a nurse transitional care coordinator used ADT alerts during discharge follow-up, but they did not use the high-needs report in their calls. After gaining access to a list of the practice roster with flags for chronic conditions, we performed K-means clustering on a subset of patients with diabetes, identifying four subgroups (1) persons with comorbid substance use disorder, (2) chronic kidney disease or (3) language barriers, and (4) persons without those comorbidities with high medical and/or social needs (31). K-means clustering is a method that groups individuals into clusters based on how similar they are to each other. In healthcare settings, the K-means algorithm can help identify patient subgroups with similar needs so interventions can be better tailored to meet those needs (32). While we chose not to offer standardized care plans based on these groupings, we incorporated these insights into ETCA design to help nurses personalize outreach and care planning. Annual guided reflections with practice partners further enabled the resource team to closely monitor the implementation process, even when they were not actively participating in the intervention. This iterative, asynchronous process allowed ongoing experience to shape further adaptations and continued engagement of all stakeholders, consistent with the DAP framework.