The quality of the patient-provider relationship is strongly linked to patients’ increased adherence and better health outcomes [1-3]. Patient-provider communication requires exchanging accurate and relevant information to better understand patients and their preferences and context [4,5]. However, provider instructions are complex and not communicated adequately to patients [3,6] as health care providers often lack the time or communication training [7]. In turn, patients have difficulties recalling crucial information (eg, their adherence to taking medication regularly), impeding providers’ ability to quickly assess their medical condition and derive actions [8,9]. Therefore, adherence and health outcomes are often subpar—especially in people with chronic conditions [6,10,11]. Due to its centrality, improving patient-provider communication is a topic of continued interest in medical research.

Over the last decades, the understanding of good patient-provider communication has evolved. Historically, providers have possessed most of the power in patient-provider communication [12]. These power dynamics mainly occurred due to the significant knowledge difference between providers and patients [13]. The resulting paternalistic model, in which the providers made all the decisions, led to poor adherence and increased health care costs [14,15]. Newer approaches such as shared decision-making and patient-centered care ask providers to adopt more inclusive methods focusing on the collaborative nature of patient-provider communication [4,5,16-18].

Patient-centered care proposes a holistic clinical method that centers on the patients and their preferences and contexts [4,5,19]. It is defined as “respectful of and responsive to individual patient preferences, needs, and values and ensuring that patient values guide all clinical decisions” and is 1 of 6 key elements of high-quality care [16]. The clinical method of patient-centered medicine by Stewart et al [4] is among the most frequently used frameworks. It proposes a communication approach that suggests looking beyond a patient’s acute problem and into their history and context (Figure 1). The framework suggests (1) exploring health, disease, and the illness experience; (2) understanding the whole person; and (3) finding common ground to (4) enhance the patient-clinician relationship. In the following paragraphs, we describe each of those dimensions.

Exploring health, disease, and the illness experience highlights the importance of understanding the patient’s experience. People have different interpretations and unique experiences of health and illness. Someone with an asymptomatic disease may perceive themselves as healthy, whereas others feel ill without having a disease. Providers should seek to understand and support the patient’s view of their situation and their experience of health and illness by listening to their concerns and feelings [4,20]

Understanding the whole person focuses on understanding the patient’s proximal and distal context [4]. It enables meaningful conversations about the illness and treatment options as providers know their patients as people (eg, what is currently important to them) [20]. Younger patients might struggle more with a diagnosis than older people. A patient’s relationships, work, education, lifestyle, and culture play a significant role in their treatment.

Patients and providers work together to find common ground regarding the problems and priorities, the goals of the therapy, and the roles of the patient and the provider [4]. It stresses the emotional engagement with the patient and the genuinely collaborative aspect of finding common ground to arrive at mutual decisions [4].

Enhancing the patient-clinician relationship is the goal of every encounter in patient-centered medicine [4]. Health care becomes genuinely patient centered through an integrated understanding of the patient’s experience, understanding them as a whole person, and mutual decisions [20]. The framework by Stewart et al [4] conceptualizes patient-provider communication to achieve patient-centered care. It proposes a mindset that places patients at the center of clinical practice.

This shift in the mindset has attracted increasing interest in health informatics research that studies the effect of technology on patient-provider communication [21-27]. In face-to-face consultations, patient data generated on mobile health (mHealth) apps become increasingly important as they allow patients and providers to gain deeper insights into patients’ routines and adherence to therapy plans. Patient-generated data are health-related data gathered or created by patients, usually through wearables and mHealth [28]. Studies in health informatics and related fields have demonstrated the potential of patient-generated data to increase patient-provider communication [22,29,30]. mHealth allows patients to generate abundant health information, such as dietary patterns, emotional conditions, or objective measures such as blood pressure [22,23,31]. These data allow for insights into the patient’s health experience and journey unlike ever before [32,33].

mHealth-supported approaches have significantly improved patient-provider communication, adherence, and health outcomes in chronic care [34-36]. Studies have shown how these patient-generated data allow patients and providers to engage in collaborative sensemaking that improves decision-making [23,30,31,37]. It allows for deeper discussions about personal values [14,16] and improves patients’ understanding of their condition and treatment [38]. Furthermore, introducing patient-generated data into consultations affects the role dynamics of therapeutic sessions [34]. For example, sharing clinical notes shifts power in the patient-provider relationship [36]. Other studies report how sharing patient-generated data through mHealth leads to greater disclosure and better communication in consultations, resulting in better health outcomes [30,31].

These insights have been validated for different age groups [30,39,40] and chronic conditions (eg, chronic kidney disease) [21,25,35,41]. This previous research shows the positive impact of introducing patient-generated data into consultations on adherence and health outcomes [35,36]. For example, Vitger et al [35] describe the positive impact of generating data on a smartphone app on patient activation, communication confidence, and preparedness for decision-making in patients with schizophrenia.

While existing research agrees on the vital role of patient-generated data in patient-provider communication [19,21,23], significant obstacles remain to leverage their potential. So far, structural and regulatory barriers have slowed advances [42,43]. mHealth apps seldom integrate with the provider’s workflow, leading to a fragmentation of health data [44,45]. More importantly, Cozad et al [46] found that only a few mHealth apps engage and activate patients to participate in patient-centered care. Finally, most studies report on the positive effects of patient-generated data on patient-provider communication, but they often fail to investigate the communication behaviors that use patient-generated data. Accordingly, patient-provider communication remains a black box that receives patient-generated data as input and creates better communication as output (Figure 2). Little to no research focuses on the design of systems that (1) integrate patient-generated data into the provider’s workflow and (2) use these data in consultations to enhance patient-provider communication. This study aimed to address this research gap by designing and evaluating an integrated digital health system that enhances the patient-provider communication.

This study addressed the research gap described previously by developing PatientHub in a design science research (DSR) approach to enhance the patient-clinician relationship [47,48]. DSR is a suitable approach as it systematically solves important general problems and generates new knowledge in the form of design principles, theoretical models, approaches, and impacts of technology use [49]. DSR proposes to ground a solution’s design in existing knowledge and theories, so-called kernel theories, to justify design decisions [50]. Due to these properties, design science is increasingly applied in medical informatics to study emerging technologies [48,51,52].

To address the research gap, we (1) designed PatientHub and (2) studied its impact on patient-provider communication by adopting the patient-centered care framework proposed by Stewart et al [4] as our kernel theory. PatientHub is an integrated digital health tool that introduces patient-generated data into consultations. We build on the strong correlation between patient-generated data and improved patient-provider communication established in recent work [34-36]. While the patient-centered care framework offers a holistic foundation for improving patient-provider communication, it lacks a clear operationalization of the 3 dimensions that offer mHealth designers and health care providers guidance on implementing patient-centered care. Specifically, it is unclear how to integrate patient-generated data into the consultation process and how patients and providers use them for patient-centered care. Accordingly, we formulated a design goal and several subgoals in line with our kernel theory (Figure 3 [4]).

The following sections present the PatientHub design, our field study approach, and the data analysis method.

The project team designed PatientHub in 4 iterations. Our project team included a medical informatics company, a health institute specializing in chronic care, and 2 research institutions.

PatientHub’s design is grounded in the patient-centered care framework [4] and leverages existing design knowledge [21-23,34,39,53]. Over the 3 preliminary iterations, we continuously evaluated and improved the design. We tested the first design iteration by applying the think-aloud method with 7 participants acting as patients [54]. After refining the design in the second iteration, a focus group of 5 domain experts from software development, medicine, and research evaluated the revised PatientHub. In the third iteration, 3 health care providers and 5 patients evaluated the design in role-plays of consultations. This paper reports on the field study evaluating the prototype with actual patients and physicians.

PatientHub aims to enhance patient-provider communication by integrating patient-generated data into the consultation. It consists of a patient app, where patients generate data, and a consultation app, where patient-generated data provide a foundation for discussion in the consultation. In the following section, we describe the design implementation using a scenario and screenshots of PatientHub. It represents a potential user story of PatientHub during the field study.

John is a patient at Laura’s clinic struggling with obesity. Last week, Laura proposed that John try PatientHub to help them advance John’s journey to better health. In the patient app on his smartphone, John tracked his dietary and activity habits in daily notes in a digital journal, filled out a general health questionnaire, and selected favorites for behavior change interventions as part of a 1-week preparation phase.

In the initial consultation, Laura and John review the journal entries, questionnaire answers, and intervention favorites in the consultation app on a tablet. The consultation app consists of 4 screens: goal setting, defining dietary and activity interventions, planning, and closing. For goal setting, John and Laura discuss target weight and therapy duration with a tablet-based visualization using sliders (Figure 4, left). As the visualization relates weight loss and duration to each other, they can discuss healthy weight loss and set realistic goals. To define dietary and activity interventions, John liked 3 interventions he would like to explore. John and Laura can discuss additional interventions from a list of obesity-friendly interventions (Figure 4, right; selected by medical professionals).

To create an intervention, they specify a name, a description (eg, 8000 steps per day), a duration (if applicable), recurrence (ie, regular or irregular), and preferred days and times (Figure 5, left). The consultation app allows them to discuss the interventions to arrive at a patient-centered therapy plan considering John’s specific context. In planning, John and Laura see an overview of the interventions in calendar form (Figure 5, right). They can adjust the therapy plan if necessary (eg, move an intervention from Monday to Tuesday). Once the therapy plan is finalized, all information is automatically shared with John on the patient app.

John tracks his health journey in PatientHub’s journal for 2 weeks. He tracks his adherence to the therapy plan and his experience while executing it. John sees each intervention in the calendar, where open interventions are grayed out and become colored once completed (Figure 6, left). For example, John had to limit carbohydrate-dense foods today. He clicks on the gray task icon to create a task-specific entry. He marks the task as completed and sets his emotional state to medium as he missed out on dessert today. John then uploads a picture of his lunch and writes a note (Figure 6, middle). He can now review his entries in his journal (Figure 6, right). John carried out the therapy plan and kept his digital journal during the 2-week implementation phase leading up to the follow-up consultation with Laura.

The follow-up consultation focuses on reviewing John’s journal. Laura and John apply different filters to the journal entries, such as task type (ie, diet, activity, and daily note), emotions, execution, and media type (Figure 7, left). This way, they can review interventions that were not completed or completed but not enjoyed by John (Figure 7, right). Through this discussion, they identify opportunities to improve the therapy plan and adherence. They adjust interventions as in the initial consultation by going through diet, activity, and planning before closing the consultation. Again, all data are shared across the PatientHub apps and the loop between consultations is closed. John enters a new implementation phase where he records his progress, which he will review again with Laura.

The field study included 28 patients and 6 health care providers (ie, physicians in this study). Only 4% (1/28) of the patients dropped out after the initial consultation (patient 24), whereas all physicians completed the 3-week study. Before the field study, we conducted a pretest in December 2022 with 2 participants to uncover and resolve software bugs and flaws in the study design. The actual field study took place in 2 clinics in Switzerland in 2 rounds between January 2023 and April 2023 for logistic reasons. The first round was conducted in January 2023 with 18 patients and 4 physicians, and the second round started in March 2023 involving 10 patients and 2 physicians. We maintained the same study design and evaluated the same prototype.

To evaluate the effects of PatientHub on patient-provider communication in the most realistic setting, the inclusion criteria for patients were (1) age of ≥18 years, (2) a BMI of >25 kg/m² or specific medical indications for weight loss (eg, diabetes), (3) ability to communicate in written and verbal German, and (4) ownership of a computer and smartphone and adequate handling of both. Physicians were selected through personal contacts and had to meet the following inclusion criteria: (1) licensed physicians in Switzerland, (2) experience counseling patients with overweight, (3) ability to speak German, and (4) familiar with computers and tablets in consultations. Tables 1 and 2 present the demographic data of the participating patients and physicians, respectively.

The framework for evaluation in DSR proposes a continuum from formative to summative evaluation [55]. The framework offers 4 evaluation strategies, from which this study adopted the “human risk and effectiveness” strategy as the addressed problem is social and user centered [55]. Our analysis consisted of summative and formative elements as the study aimed to improve on the process under evaluation (ie, patient-provider communication). Accordingly, we applied deductive and inductive coding methods in 3 steps for the data analysis, as depicted in Figure 9.

To study the effect of PatientHub on patient-provider communication, we analyzed individual interviews with patients and physicians and 54 video recordings of face-to-face initial and follow-up consultations (27 recordings each). This three-step analysis allowed us to assess (1) whether and (2) how PatientHub enhances patient-provider communication (interviews) and to (3) observe recurring behaviors when engaging with patient-generated data in consultations (video recordings). The research team consisted of a graduate student with a medical and information systems background (coding author), a PhD student in digital health (lead author), and 2 senior researchers in DSR.

First, we assessed the design goal attainment “enhance patient-provider relationship” by analyzing the interviewees’ accounts regarding their perception of their relationship with the physicians (summative). After determining whether PatientHub had enhanced the patient-provider relationship, the coding author conducted a thematic analysis of the interviews [56]. This allowed us to identify the aspects that characterize high-quality patient-provider communication when using PatientHub (formative). The coding author created in vivo codes focusing on the effects on patient-provider communication attributed to patient-generated data in the consultation. The lead author conducted quality assurance by reviewing and revising the codes. In an iterative process, the lead and coding authors then grouped the codes into 4 characteristics of high-quality patient-provider communication: personalization, actionability, trustworthiness, and equality.

Second, we assessed the attainment of our subgoals “exploring experience,” “understanding person,” and “common ground” by applying a mixed deductive and inductive analysis [57,58]. We developed an initial coding scheme from related work on our kernel theory, patient-centered care [4,20]. We complemented the coding scheme with codes derived from the literature on patient-provider communication [22,23,38,39,59]. Again, the coding author created in vivo codes to capture emerging phenomena. She coded approximately 20% of the interviews before discussing the results with the lead author to refine the coding strategy. The coding author then finished coding all interviews. Finally, the lead author reviewed and refined the coding by discussing discrepancies with the coding author. In a workshop, the author group synthesized how PatientHub improved each dimension of patient-centered care.

Third, we deductively analyzed the video recordings regarding emerging behaviors [58]. The coding author analyzed the recordings based on the coding scheme applied and refined in the interview analysis. The coding scheme sensitized us to patient-centered behaviors enabled through PatientHub. During the coding of the videos, the coding scheme was expanded with in vivo codes to include emerging phenomena [58]. The video recordings were analyzed in 2 rounds on-site at one of the clinics (to ensure data privacy). We extracted emerging behaviors from the first coding round by drawing sequential processes. Initial drafts of these processes were discussed in a workshop including the author group and members from the project consortium. In the second coding round, we iterated on all video recordings based on the identified behaviors to refine our understanding and transcribed relevant sequences from the recordings. This step allowed us to formalize the 3 communication behaviors described in the Results section.

While the ethics committee of the canton of Zurich confirmed that this study is not subject to the Swiss Human Research Act (BASEC-Nr. Req-2018-00847), we still decided to obtain written consent from all patients and physicians before the field study. The informed consent form educated participants about their rights and responsibilities, data use, and privacy measures following the World Medical Association Declaration of Helsinki [60]. The participants had the possibility to opt-out at any time during the study. The form also informed the participants’ data will be gathered deidentified by assigning each participant with an identifier (ie, patient 1 to patient 28 and provider 1 to provider 6). Multimedia Appendix 5 provides the authors’ positionality statement to give the readers a better understanding of the authors’ backgrounds.

The participating clinics selected patients using a purposeful sampling approach [61]. The compensation for patients was CHF 50 (approximately US $60) and a raffle ticket for a restaurant voucher worth CHF 200 (US $230.15).

In this section, we present our findings. First, we examine the attainment of the design goal (ie, enhance the patient-provider relationship) and discuss the 4 characteristics of high-quality patient-provider communication identified in the analysis. We then examine how patients and providers perceived patient-centered care in these consultations (ie, subgoals). Finally, we introduce 3 emerging communication behaviors when engaging with patient-generated data.

Our analysis elicited the 4 characteristics of high-quality patient-provider communication. In the following sections, we explore the process of enhancing patient-provider communication by discussing the 3 dimensions of patient-centered care (ie, subgoals)

Supporting patient-provider communication using patient-generated data is a growing topic of interest in health informatics and related fields. However, the existing literature often overlooks the processes (ie, communication behaviors) when enhancing patient-provider communication using patient-generated data. More importantly, existing mHealth apps are seldom integrated into the provider’s workflow and do not sufficiently engage and activate patients in patient-centered care [44-46]. Through a design science approach, we explored how to design an integrated digital health system and its impact on the communication behaviors of 27 patients and 6 providers.

Previous work has demonstrated the benefits of integrating patient-generated data into consultations [32,35,38,45,63,64]. We expand on this work by studying the effect of patient-generated data on patient-provider communication behaviors. As argued in the Introduction section, current research focuses on the input (patient-generated data) and output (enhanced patient-provider communication; Figure 2). Our analysis elicited 3 emerging behaviors that open the black box of patient-provider communication. On the basis of our theoretical foundation and empirical findings, we propose patient-centered communication in 3 design principles to operationalize the patient-centered care framework.

The emotion labeling behavior highlights how the input is initially processed. Patient-generated data enable patients and providers to reflect on patients’ emotions related to their health. For example, patients speak about their insecurities when engaging in physical activities in public due to being overweight. Prompted by photos, emojis, and text, patients and providers explore patients’ unique experience with their health and disease (“exploring experience” subgoal). Furthermore, emotion labeling uncovers obstacles in the patients’ journeys to better health. As described in the Results section, journal entries allowed patient 03 and provider 01 to uncover and discuss the reason for the patient’s nonadherence. Patient-generated data allowed the patient and provider to understand the patient as a whole person and their proximal context (“understanding person” subgoal).

Existing research has shown how visualizations of health data increase health literacy in patients [38]. Our results show how patient-generated data are used to not only educate patients but also collaboratively generate new insights. We argue that patient-generated data enhance patient-provider communication by stimulating the exploration of a patient’s experience and context. Providers can usually only scratch the surface of a patient’s story in consultations. PatientHub introduces them to their patients’ world, allowing them to dive below the surface and gain an integrated understanding. The patients themselves also profit from their data. Our results show that generating data often initiates self-reflection that fosters patients’ understanding of themselves. Therefore, patient-provider communication becomes more patient centric. Patients and providers explore the patient’s experience and context to find common ground regarding problems and goals.

The expectation decelerating behavior processes patient-generated data differently. Instead of beginning with a patient’s experience or context, patients and providers used patient-generated data to assess problems or goals. Our results show 2 areas in which patient-generated data decelerated patients and providers: setting goals and developing therapy plans. Providers referred to these data in the former to decelerate the patients’ ambitions. They suggested either more long-term planning or a more achievable intermediate goal. When developing therapy plans, patients and providers resisted the urge to add more or exchange interventions too quickly due to patient-generated data (“common ground” subgoal). As seen in the Results section, patient 01 could successfully argue his standpoint on keeping intermittent fasting in his therapy plan due to his data. The data allowed them to assess the therapy plan’s effectiveness, resulting in the mutual decision not to adapt it if the effectiveness persisted. Existing research demonstrates the value of patient-generated data for problem-solving and decision-making [23,64]. While we agree with their findings, the saying “It’s a marathon, not a sprint” highlights the importance of expectation decelerating. Our results show how patient-generated data are applied in consultations to resist the urge to hurry long-term behavior change. Accordingly, expectation decelerating might counteract the common pressure of a quantified self to always strive for more [65]. Quantified humans use data to continuously find inefficiencies and improve on those, often resulting in unhealthy pressure. Instead of expanding the therapy plan, patients and providers could assess the effectiveness of the current plan in the follow-up consultation. They had a reliable foundation for decision-making instead of hampering long-term success with impulsive actions. They defined achievable targets with a personalized therapy plan tailored to the patient. Therefore, patients might have more endurance in their marathon race to change their lifestyle.

Decision ping-pong reflects on how patient-generated data affect the dynamics of patient-provider communication. Existing research in medicine and human-computer interaction postulates the importance of shared decision-making [4,17,21,23,53]. Our results expand on these insights by examining how patient-generated data include the patient in decision-making as an expert and how this changes the consultation dynamics (“common ground” subgoal). Many patients noticed how their position had changed in the initial consultation compared with their previous experience. We argue that introducing patient-generated data goes beyond facilitating informed decision-making. With these data, patients provide “proof” of their adherence, for example, through pictures. Moreover, they have a much deeper knowledge of PatientHub’s content. It contains patient-generated data and evokes memories that might not be documented in the system. Equipped with this knowledge, patients become experts in their own domain: their health journey.

In shared decision-making, the implicit understanding is that providers assess the patient’s state and offer options [17]. Existing research often sees patient-generated data as crucial information for providers to offer better health services [22,64]. Studies explore ways to design technology to make large amounts of patient-generated data consumable for providers [23]. Implicitly, this focus places the responsibility for data interpretation on the providers. However, we believe that patient-generated data could partially relieve providers from this burden. Patient-centered care emphasizes mutual decisions; therefore, patients should carry this burden with their providers.

Our results show that most patients willingly accepted such a role as they felt taken more seriously, could defend their position, and perceived having control over the decision-making process. Integrating patient-generated data into the consultation empowers patients to assume a more active role in their health care. We argue that providers must not be solely responsible as our results highlight how patients become experts in their health journey. Instead, the responsibility is shared between patients and providers according to their expertise. While providers assess the data against their medical knowledge, patients interpret the data in the context of their lives and experiences. Therefore, their communication becomes truly “respectful of and responsive to individual patient preferences, needs, and values and ensuring that patient values guide all clinical decisions” [16].

This study evaluated communication behaviors that emerge when introducing patient-generated data into patient-provider communication. We studied how these behaviors in patient-provider communication actualize the potential of patient-generated data to increase patient-centeredness. Our analysis uncovered 3 communication behaviors in medical consultations when using PatientHub. Furthermore, we demonstrated how enhanced patient-provider communication is necessary for patient-centered care. On the basis of our findings, we believe that this study contributes to research in 2 ways. First, we emphasize the value of patient-provider communication. The identified behaviors demonstrate how data-supported patient-provider communication creates value, not the technology and data themselves. Second, the behaviors offer actionable insights into implementing patient-centered care.

However, this study does not come without limitations. First, while we evaluated PatientHub in the most realistic setting, its applicability in the real world depends on regulatory and security frameworks. Second, the generalizability could be further increased with a larger sample size and a randomized controlled trial. Future research could also investigate how patient-generated data empower patients individually to study the changing role dynamics in medical consultations.