Figure 1 depicts the eRIKA process for ePrescribing and medication safety checks as originally designed and served as the basis for the LL activities. It is based on centrally stored, individual electronic medication plans accessible to health care providers and patients. Each plan is intended to be continuously updated using ePrescriptions; prescription and dispensation data from statutory health insurance claims data; and information on over-the-counter (OTC) drugs provided by PHs and patients. This eRIKA medication plan differs from the national TI medication plan in 2 ways: it draws on additional data sources (claims data and OTC information) and is potentially available to a wider audience, as the TI medication plan is limited to patients receiving at least three prescribed medications.

The eRIKA system contains patient data provided by statutory health insurance providers and supports the electronic review of prescriptions to detect preventable risks. It also provides PHs with detailed information to support effective pharmaceutical counseling. CDSS functions, such as medication interaction alerts and dosage checks, are integrated to enhance prescribing safety. Additionally, the system includes digitally assisted communication between physicians and PHs, as well as drug traceability to enable the prompt identification and resolution of safety issues.

The system comprised components at different stages of technical maturity when the LL co-design process began. The CDSS was implemented as a functional prototype: it was technically operable within the project environment, but its functionalities and interface design were subject to iterative refinement through the co-design process. The communication functions, which would enable structured digital exchange between general practices and community pharmacies, were planned for future implementation. The ePrescription and electronic medication plan were operational prototypes. Both were fully implemented and operational by January 2024 as part of the federal rollout of ePrescriptions. The intervention as a whole depends on the national TI as its underlying infrastructure.

Figure 2 shows the planned technical integration of the software as an independent software component within the existing national TI operated by gematik. Key actors depicted are the GP, community pharmacy, patient, and statutory health insurance. Security-relevant components of the national health TI shown are the ePrescription, the Health ID (Gesundheits-ID), the TI authenticator, and the interface connecting eRIKA to the broader TI ecosystem. Figure 2 reflects the status of the operational prototype at the time of the LL co-design process. Not all depicted components were fully implemented during the study period, and certain functions represent planned integrations that depended on external systems within the national health TI.

To develop, implement, and evaluate the new form of care, the project was divided into 3 stages. The first stage, reported in this manuscript, was the LL [13]. The LL was conducted between November 2022 and September 2024, during which the software interface and the processes between GP practices and pharmacies were iteratively developed and refined. This stage was followed by 2 implementation and evaluation phases. In phase 1, the intervention was initially tested on a smaller scale in a cluster-randomized controlled trial [14]. Data collection for phase 1 has been completed, and analysis is ongoing. In phase 2, the intervention was expanded and implemented on a broader scale in a quasi-experimental study [14]. Data collection for phase 2 has likewise been completed, and analysis is ongoing. The overall project ran from October 2022 to September 2026.

The LL conducted as part of the eRIKA study was approved by the Ethics Committee of the University of Wuppertal (approval number SK/AE 220906). Before participating, all participants provided written informed consent covering both their participation in the workshop sessions and the audio recording of these sessions. Participation was voluntary, and participants were informed of their right to withdraw at any time without consequence. Participants received financial compensation for their time and travel (€276 [US $310] net per practice/pharmacy team per sprint or €212 [US $238] net if only 1 person participated per sprint).

To protect participants’ privacy and confidentiality, transcripts were anonymized using codes. Data were handled in accordance with applicable data protection regulations, including the GDPR (General Data Protection Regulation).

LLs can be used as a tool within the co-design process, providing a collaborative platform for active discussion and iterative refinement of the planned intervention with potential users. In this context, co-design refers to the active involvement of end users, such as GPs, PHs, and patients, together with developers and researchers throughout the development or implementation process. The specific goal of this project was to collaboratively refine the details of the care processes to align them closely with user needs, expectations, and real-world practices and to define concrete, practical steps for effective implementation within participating practices and pharmacies.

The methodology applied in this project followed a structured, iterative co-design process divided into 4 consecutive development “sprints” following an onboarding meeting. A sprint refers to a defined development cycle consisting of several sequential steps: (1) preparation and presentation, (2) interest-holder workshops, (3) documentation and ranking of change requests, and (4) negotiation meetings (for sprints 1 and 2).

The onboarding meeting served to introduce the LL concept and prepare participants for the co-design process. Before sprint 1, a 1-hour meeting was held for participants from 2 of the 3 regions: Westphalia-Lippe (region 1) and Berlin (region 2). The third region, Saarland, joined the study at a later stage, and a separate onboarding meeting was held for its participants.

The interest-holder workshops took place at the premises of KVWL (Kassenärztliche Vereinigung Westfalen-Lippe; region 1), AOK Nordost (region 2), and the software provider based in Saarland (region 3). Each sprint addressed a different topic, as described below. All participating organizations were directly involved in the project, which may have discouraged the open expression of critical or dissenting views. This environmental framing represents a form of contextual bias that cannot be retrospectively controlled for.

Online negotiation meetings were held during sprints 1 and 2 to discuss change requests arising from the workshops. These meetings involved workshop participants and members of the core project team. Members of the core project team first ranked the requests according to 3 feasibility dimensions: temporal feasibility (realizable within the project timeline), technical feasibility (within the IT architecture), and financial feasibility (within the project budget). Each dimension was rated on a scale from 0 to 3, with 3 indicating full feasibility and 0 indicating no feasibility, yielding a composite score for each change request. The resulting scores indicated whether the core team considered a request immediately implementable, potentially implementable at a later stage, or not implementable. The rationale for each assessment was discussed during the negotiation meeting, during which workshop participants could argue for a change in the assessment.

No negotiation meetings were held after sprints 3 and 4. This decision was deliberate rather than incidental. Given the nature of these sprints, which centered on live software demonstrations and hands-on system walkthroughs, representatives of the software development team were present throughout each session. This enabled real-time documentation and direct discussion of requirements identified by participants within the workshops themselves, making separate post hoc negotiation meetings unnecessary.

The LL participants comprised 4 professional groups working within the German primary care and community pharmacy system: GPs and medical assistants (MAs; Medizinische Fachangestellte) working in GP practices, as well as PHs and pharmacy technicians (PTs; Pharmazeutisch-technische Angestellte). All 4 professional groups are directly involved in the prescribing or dispensing workflows addressed by the project. Patients were not included in the LL because the focus was on the design and optimization of the electronic prescribing workflow and the associated software interfaces within GP practices and pharmacies, which are system components with which patients do not interact directly. Overall, 6 general practices, 6 pharmacies, and 1 hospital were represented across the 3 regions. Recruitment followed a convenience sampling strategy, drawing on existing professional networks and prior project collaborations in each region. In Saarland, the Saarland Association of Statutory Health Insurance Physicians (Kassenärztliche Vereinigung Saarland) recruited 2 general practices affiliated with the association as pharmacotherapy advisors. These practices, in turn, identified 1 interested community pharmacy within their professional networks. In Berlin, a regional statutory health insurance company (AOK Nordost) recruited 2 general practices with prior experience in a related project and 2 pharmacies, with support from the Berlin Pharmacists’ Association (Apothekerverband Berlin). In Westphalia-Lippe, the KVWL recruited 2 GPs who had previously participated in a project. These physicians subsequently approached pharmacies with which they had established working relationships. The participating hospital was approached by the project team based on its participation in a previous project.

The study included micro-level actors (GPs, MAs, PHs, and PTs) as LL participants. Meso-level actors were represented in the negotiation meetings and included the software provider, statutory health insurance companies (BARMER and AOK Nordost), and the KVWL. Macro-level interest-holders were not involved, reflecting both the study’s focus and the contextual realities of the German health care system. Consistent with the MRC Framework’s [8] guidance on interest-holder engagement in complex intervention development, involvement was prioritized according to relevance to the intervention’s core objectives. This study focused on end-user experience and organizational workflow integration. Additionally, in Germany’s highly fragmented health care system, the perspectives of macro-level organizations are inherently shaped by their institutional and political positions and may not reflect the diversity of views across the system.

Workshops were conducted between November 2022 and September 2024 as structured, 2-hour sessions. To ensure accurate data collection, written notes were supplemented by audio recordings. Each workshop addressed a specific component of the intervention or its current state, which was presented as a description, mock-up, or functional prototype. The “thinking aloud” technique [15] was used to elicit participants’ perceptions and identify workflow-related challenges and areas for improvement, particularly at the interface between general practices and pharmacies. The LL sessions were conducted in German and moderated by trained members of the research group (AP, IM, and BM). Other members of the research group recorded the sessions and took field notes. Data collection was organized into sprints with predefined themes and development steps, as described below. These themes and subsequent development steps were adapted iteratively in response to findings from previous sprints. Triangulation was achieved by comparing data across the 3 regions. These methods were selected to support a comprehensive process evaluation of the LL and capture participants’ experiences and perspectives throughout the co-design process.

Sprint 1 was conducted in 2 regions (regions 1 and 2) as individual 2-hour workshops (n=4 and n=7, respectively), which were audio-recorded in full. The sessions focused on generating broad, formative feedback on all components of the eRIKA system and the processes developed for physician practices and pharmacies, as well as identifying priority topics for further development. The outputs of sprint 1 directly informed a revision of the eRIKA process design, which formed the basis for sprint 2.

Sprint 2 was conducted across all 3 regions, with each session lasting 2 hours, for a total of 4 workshop sessions. In region 1, separate sessions were held for physician practices (n=2) and pharmacies, as originally planned for all regions. In regions 2 and 3, scheduling constraints required joint sessions for both settings (n=7 and n=6, respectively). The sessions addressed specific feedback on the revised version of the eRIKA process developed after sprint 1, with particular attention to practice- and pharmacy-specific workflow aspects. All sessions were audio-recorded. In region 3, content from sprint 1 was also revisited at the beginning of the session. Feedback from sprint 2 was documented in the change log and informed further software refinements before sprint 3.

Sprint 3 was conducted across all 3 regions as individual 2-hour workshops (region 1: n=3; region 2: n=5; and region 3: n=5), all of which were audio-recorded in full. The focus shifted from process design to the eRIKA software interface, including its design, usability, and acceptance, and the collection of specific improvement requests. In each session, the software provider gave a live demonstration of the practice software (remotely in 2 regions, with the project team and participants on-site, and fully on-site in 1 region). No printed or digital materials were distributed; the live demonstration served as the sole basis for discussion.

Sprint 4 was conducted 3 times across 2 regions (region 2: 2 sessions, n=4 each; and region 3: 1 session, n=3), with each session lasting 2 hours and audio-recorded in full. Sprint 4 could not be conducted in region 1 because of technical and scheduling challenges. The sessions focused on challenges in collaborative workflows between physician practices and pharmacies. Participants conducted a hands-on walkthrough of all eRIKA process steps in a live test environment, providing direct experiential feedback on interactions between the 2 settings. No additional materials were used beyond the live system.

Notes from meetings with GPs, MAs, PHs, and pharmacist assistants were analyzed to identify recurring themes and practice-relevant insights. Data collection was therefore iterative, capturing evolving user needs and adaptations to local workflows over time.

All LL sessions were audio-recorded, transcribed, anonymized, and analyzed in German. For anonymization, participants were assigned acronyms (GP, MA, PT, and PH). Participants were then assigned numerical identifiers based on the order in which they spoke. After the analysis, the results and selected quotes were translated into English for publication by native German speakers with a high level of English proficiency. The translation was assisted by the AI system Claude (Anthropic) and subsequently independently reviewed and finalized following the four-eyes principle.

This study is grounded in an interpretivist research paradigm, which assumes that knowledge is socially constructed through participants’ lived experiences, perspectives, and interactions. Data were analyzed using reflexive thematic analysis within an interpretative (Big Q) spectrum, as outlined by Braun and Clarke [16,17]. Consistent with Braun and Clarke’s [17] emphasis on methodological coherence and reflexive openness, reflexive thematic analysis was chosen over more structured approaches, such as qualitative content analysis (eg, Mayring) or codebook thematic analysis, because it offers the flexibility required to capture the dynamic, iterative nature of interest-holder feedback within an evolving co-design process. Given the exploratory character of the LL process, the diversity of data sources (eg, observational notes and negotiation meetings), and the heterogeneity of participants, an open and adaptive coding strategy was considered more appropriate than the application of a predefined category system. Furthermore, because the care process itself evolved across the 4 sprints, codes were continuously adapted to accommodate emerging topics, a process that would be constrained by a fixed codebook. Coding was carried out independently by multiple researchers (AP, EG, and CB) and subsequently discussed and refined collaboratively. Divergences were resolved through iterative discussion, consistent with the reflexive rather than reliability-oriented tradition of this approach.

In accordance with this approach, the thematic analysis was divided into 6 steps: (1) data familiarization, (2) generation of initial codes, (3) search for themes, (4) review of themes, (5) definition and naming of themes, and (6) reporting of results. The initial analysis was carried out by researchers with expertise in implementation science, qualitative methods, and eHealth interventions, ensuring methodological rigor and contextual sensitivity throughout the analytic process. The researchers who facilitated and analyzed the LL sessions had academic backgrounds in sociology and psychology. As members of the project team, they had prior knowledge of the study and its development, which may have introduced assumptions about which aspects of the co-design process were particularly salient or problematic. The researchers’ dual role as both session facilitators and data analysts created a degree of relational familiarity with participants, which may have encouraged more open disclosure but could also have influenced workshop dynamics.

After initial data familiarization and code generation (steps 1 and 2), codes and preliminary themes were compared and discussed. Where interpretations differed, the differences were discussed until consensus was reached, and code definitions were adjusted. Following steps 1 and 2, the initial codes were refined and expanded to reflect nuances in the material, particularly in response to the thematic focus of each development sprint. Coding was performed in Microsoft Excel, with regular reflexive review to ensure analytic rigor and depth. This process resulted in the following list of preliminary categories, which was used to structure all relevant data:

Based on these initial codes, data were then iteratively clustered into preliminary themes and subthemes (steps 3 and 4), aiming to reflect interest-holder needs as articulated during the LL sprints. Given the varying methodological emphasis and content across the 4 sprints, not all codes were applicable in every case, and additional codes were introduced as needed to capture emerging requirements and process adjustments. Through iterative discussion and reflexive analysis across the research team, the initial codes were progressively consolidated into 3 overarching themes—Workflow Efficiency, Technical Implementation, and A New Way of Working—each integrating multiple initial codes that shared a common conceptual focus (step 5). These themes were subsequently reported (step 6). This study was reported in accordance with the SRQR (Standards for Reporting Qualitative Research) checklist (see Multimedia Appendix 1).

The sample comprised 17 participants, including 10 from medical practices (GPs and MAs) and 7 from pharmacies (PHs and PTs; see Table 1). The same participants were invited to attend all sprints throughout the co-design process to ensure continuity of involvement. Due to scheduling constraints, not all participants were able to attend every session, but continuity was maintained at the practice and pharmacy levels.

The adequacy of the sample was assessed based on information richness rather than numerical sufficiency. The sample was considered appropriate given the depth and breadth of professional experience represented, the multiregional composition, and the iterative nature of the data collection process, which allowed emerging themes to be revisited and elaborated across successive sprints.

To illustrate the breadth of interest-holder involvement and the saturation of qualitative data across professional groups, Table 2 maps the 3 themes and 8 subthemes generated through thematic analysis to the 4 participating interest-holder groups (GP, MA, PH, and PT). All 3 themes were informed by contributions from at least three interest-holder groups, with theme 2 (Technical Implementation), specifically subtheme 2.3 (System Interoperability and Integration), being the only subtheme to which all 4 professional groups contributed. GPs provided input across all 3 themes and all 8 subthemes, reflecting their central role in the prescribing workflow and their broad engagement with both technical and organizational aspects. PHs contributed substantially to themes related to workflow efficiency and technical implementation, while their input was less prominent in discussions of roles, responsibilities, and regulatory requirements—areas predominantly raised by GPs and MAs. MAs and PTs contributed more selectively, with MAs focusing primarily on workload-related and regulatory concerns within the practice setting, and PTs focusing on communication and interoperability issues at the pharmacy counter. The distribution of contributions across groups reflects the complementary perspectives inherent in the collaborative prescribing process and supports the thematic saturation of the dataset within the scope of the LL co-design approach.

The themes identified through reflexive thematic analysis were interpreted through the lens of Normalization Process Theory (NPT) [18]. NPT provides a framework for understanding how new practices and technologies become embedded, or fail to become embedded, by focusing on 4 generative mechanisms: coherence (sense-making), cognitive participation (engagement and commitment), collective action (the work of implementation), and reflexive monitoring (appraisal of outcomes). Applied to our findings, all 4 mechanisms are reflected across the 3 themes, although with varying prominence.

The sense-making work (coherence) through which participants understand the purpose and value of a new practice is most prominently reflected in theme 2 (Technical Implementation), where participants repeatedly questioned the added value of the intervention relative to existing tools, the reliability of the medication history as a basis for clinical decisions, and the system’s interoperability with their existing software environments. However, coherence concerns also permeate theme 1, where participants’ skepticism about alert overload and system complexity suggests that the value proposition had not yet been fully established across all professional groups.

The relational work through which key actors are enrolled and committed to the new practice (cognitive participation) is evident in the structural decisions made throughout the LL process itself: the deliberate inclusion of all 4 professional groups across 3 regions, the involvement of the software provider as an active participant in sprints 3 and 4, and the separate facilitation of practice and pharmacy sessions in region 1 to enable setting-specific engagement. Theme 3 further reflects cognitive participation in participants’ negotiation of new roles and responsibilities.

The operational work of enacting the new practice in context (collective action) is captured primarily in theme 1 (Workflow Efficiency), where participants across all professional groups identified practical barriers, including cognitive load, communication inefficiencies, and the risk of workflow disruption. The sprint structure of the LL was itself a form of structured collective action, enabling participants to progressively engage with and refine the intervention across cycles.

The appraisal work through which participants assess the effects of a new practice and adjust their engagement accordingly (reflexive monitoring) is reflected across all 3 themes but is most explicit in theme 3, where participants evaluated regulatory implications, data privacy risks, and the long-term sustainability of proposed workflow changes. The LL format inherently supports reflexive monitoring, as each sprint provided a structured opportunity for participants to evaluate the revised intervention and feed their assessments back into the next development cycle.

This study examined the application of an LL methodology to iteratively co-design and refine a collaborative electronic prescribing intervention (eRIKA) for patients with polypharmacy in German primary care and community pharmacy settings. Three themes emerged from the thematic analysis: Workflow Efficiency, Technical Implementation, and A New Way of Working. The study also reflects on the practical challenges encountered in applying the LL methodology within a project plan. Interest-holder–derived insights gathered across 4 sprints were fed back into the design process through a collection of change requests, followed by a negotiation procedure. The LL aimed to facilitate the stepwise refinement of the intervention in preparation for 2 implementation studies on a small scale (10 practices and 10 pharmacies) and a medium scale (116-233 practices and corresponding pharmacies).

The LL findings encompassed both expected and unexpected subthemes. Concerns about workload, system interoperability, roles and responsibilities, and regulatory barriers to implementation are not specific to the intervention but have been reported for a wide variety of health IT implementations [1,19]. Alert fatigue and information overload, in particular, have been discussed previously in CDSS interventions [20]. The LL findings suggest that participants wanted filtering mechanisms or prioritization strategies to manage the volume of available information and messages. This aligns with the notion that new information technologies perpetuate a self-reinforcing cycle of information overload, underscoring the critical need to develop and refine effective filtering strategies to support better decision-making [21].

Moreover, regarding interprofessional cooperation between physicians and PHs, the findings align with previous research indicating a shared desire for a predefined, clear, and straightforward method of communication between the 2 professions [22].

Less anticipated were the substantial differences in views between GPs and PHs regarding the data PHs would need to conduct effective medication safety reviews. These differences reflect deeper interprofessional asymmetries in the understanding of the PH’s role in collaborative medication management workflows.

The LL discussions also revealed certain double binds. One concerned the use of direct digital communication (see subtheme 1.2), which was considered useful if implemented as a push messaging service, preferably accessible at the point of sale in the pharmacy when the patient is present. By contrast, clear concerns were raised about information overload and the risk of creating unrealistic expectations of immediate responses. Another study focusing on nurses and physicians highlighted the need for institutional guidance on appropriate text messaging between professions, attributing this need to fundamentally differing perspectives and expectations between the 2 groups [23]. Based on the available results, some form of guidance would also be necessary in this case, although it should be developed at the national level.

Another double bind relates to the role of patients in the process (see subthemes 2.1 and 2.2). Both GPs and PHs supported an active role for patients in which they could add information, particularly on OTC medications, including those purchased from online pharmacies. At the same time, this possibility raised concerns about the reliability and integrity of such data. By their nature, these double binds cannot be readily resolved at a purely functional level but require a more in-depth design of the process elements concerned.

While eRIKA did not aim to systematically evaluate the LL methodology itself, the experiences gained allow for selected methodological reflections. LLs offer insights into the refinement and contextual adaptation of complex interventions. This study illustrates how such a participatory approach can support iterative development and reveal barriers to the planned intervention. The structured co-design process enabled the identification of practical barriers and the direct translation of interest-holder feedback into requirements for software modifications. The LL also allowed workflow issues to be identified at an appropriate stage of the intervention development cycle, underscoring the importance of timing in co-design efforts.

Additionally, integrating workshops, negotiation meetings, and prototype testing contributed to a multidimensional understanding of user needs that would have been difficult to capture through other evaluation approaches, such as post hoc interviews, standardized surveys, or 1-time usability tests. The project also aligns with key elements of the MRC Framework for the development and evaluation of complex interventions, including interest-holder engagement, identification of uncertainties, and iterative refinement of the intervention within its real-world context [8].

Regarding interest-holder integration, the MRC Framework recommends including representatives at the micro, meso, and macro levels. This study covered the micro level (users in medical practices and pharmacies) and the meso level (IT company). However, the macro level could not be included due to recruitment difficulties. Relevant macro-level actors would have included pharmacy associations, medical associations, vendors of practice software systems, gematik (the operator of the German health TI), and political decision makers. These actors were only loosely affiliated with the project or not affiliated at all, limiting opportunities for their effective inclusion. This lack of macro-level involvement is reflected in the results. For example, participants often expressed a desire to integrate the new software into legacy practice software systems. However, this could not be implemented because of a lack of open interfaces, which are not strongly mandated by regulation.

Additionally, participant feedback indicated that the LL concept itself was not always well understood. Unfamiliar terminology and a lack of clarity about the purpose of their involvement occasionally led to confusion or resistance. After follow-up conversations and clearer explanations of the co-design process, participants reported feeling better informed and more motivated to contribute, which reinforced the participatory value of the LL format and its potential to promote engagement throughout the development cycle. This suggests that LL co-design processes should prioritize clear communication of expectations and the link between participation and tangible outcomes.

Despite various strengths, the application of the LL approach in the eRIKA project also revealed challenges. While the process facilitated rapid iteration, it lacked standardized indicators for evaluating implementation success. This aligns with existing criticism of LLs, which often rely on qualitative feedback but rarely provide quantifiable outcome data. Although one could argue that this is a goal of the approach, Zipfel et al [24] and Kim et al [25] have also noted that the field still lacks validated frameworks to systematically assess the impact of LLs beyond narrative or anecdotal results.

Despite the insights gained, this study has several limitations.

First, it reports on the formative stage of the eRIKA project and should be interpreted accordingly. The findings reflect participant-generated feedback, preferences, and perceived barriers within the specific co-design process rather than a broader, systematic view of the topics under study.

Second, the sampling strategy relied on convenience sampling through professional networks and prior project collaborations, which may have introduced selection bias by overrepresenting digitally engaged and organizationally well-connected participants. This limits the transferability of the findings to care settings with lower levels of digital maturity or weaker interprofessional ties.

Third, the study was limited to 4 thematic sprints conducted over a defined project period, restricting the ability to observe long-term adoption patterns or assess the sustainability of implemented changes beyond the co-design phase. Therefore, participant engagement and system acceptance under routine conditions remain unknown.

Fourth, the LL sessions focused primarily on intervention-level optimization, while implementation strategies to support sustained maintenance were only marginally addressed. This may limit the practical applicability of the findings to sites seeking to scale the intervention beyond the project context.

Fifth, the researchers who facilitated and analyzed the sessions were members of the project team, and their prior knowledge of the intervention’s development history may have influenced the interpretation of the results based on their personal experiences, potentially leading to an underemphasis of critical or contradictory findings.

Finally, the co-design process engaged primarily micro-level interest-holders who are directly involved in the day-to-day prescribing and dispensing workflow. Macro-level interest-holders, including health care policy makers, regulatory authorities, and representatives of the TI, were not systematically integrated into the LLs. This absence has relevant implications for interpreting the study as a comprehensive LL implementation.

By applying the LL approach, barriers, obstacles, and user requirements regarding the planned intervention could be collected systematically. These findings relate not only to technical implementation but also to challenges in interprofessional collaboration, such as communication between medical practices and pharmacies, as well as questions of responsibilities and areas of accountability. The results highlight the critical importance of involving end users in the design of health care solutions. Future research on eHealth involving multiple professions must take these findings into account and incorporate them into the research process so that potential barriers to implementation can be addressed at an early stage.

Beyond the descriptive findings, this study makes several contributions to the literature on digital health co-design and eHealth implementation. The interprofessional tensions surrounding the sharing of patient data with PHs represent a substantive finding that goes beyond usability concerns. GPs and PHs held divergent expectations regarding what data should be shared and at what level of detail, reflecting deeper professional boundaries. These tensions are not merely technical interoperability problems but also reflect boundaries between professional groups whose responsibilities are being redefined. This finding aligns with NPT’s concept of coherence: the intervention cannot be normalized until a shared understanding of its purpose and the appropriate scope of data sharing is negotiated across professional groups. Second, the double bind around push communication constitutes a theoretically relevant paradox for eHealth design. Participants simultaneously requested proactive push notifications to improve responsiveness and expressed concern that such notifications would add to an already unsustainable workload. Additionally, this study offers transferable lessons about the application of the LL methodology within a fragmented national health IT environment. Germany’s decentralized health care system, characterized by heterogeneous practice management systems, regionally variable digital maturity, and a nationally mandated but recently introduced TI, created structural constraints that shaped the co-design process at every stage. Variation in system landscapes across practices and the dual dependency on both local practice software and national TI components meant that participant feedback could not always be translated into universally applicable design solutions. This experience highlights a fundamental challenge for LLs in fragmented health IT environments: co-design processes that work well at the level of individual workflows may encounter systemic barriers at the infrastructure level that lie outside the scope of participatory design. Future LL implementations in comparable contexts should consider explicitly mapping national infrastructure dependencies at the outset and involving macro-level IT governance interest-holders, such as gematik or national interoperability bodies, alongside end users.

This study demonstrates that LLs can support the refinement of digital health interventions under everyday conditions. To fully realize this potential, several improvements are necessary. Future implementations should aim for broader participant inclusion across varying levels of digital literacy and levels of routine practice with digital tools to reduce selection bias. Linking LL activities to established methodological frameworks, such as the MRC Framework, could enhance consistency and comparability across projects. In the context of the MRC Framework, particular consideration should be given to involving interest-holders at all levels, including the political (macro) level. The LL showed that, given the particular challenges surrounding digitalization in Germany, user involvement alone is insufficient. Successful development and integration of eHealth solutions into care delivery also require the active participation of policy makers and decision makers. Future research should therefore account for this broader context and ensure that the micro, meso, and macro levels are all considered during development and implementation.