This study was part of a larger intervention study (currently under review) that included 4 primary care practices, located in Michigan (n=1), Florida (n=1), and New York (n=2), all with some level of behavioral health integration. The intervention arm (n=2 practices, 1 residency and 1 private practice) piloted the use of the Primary Care Path, a program with a patient-facing mobile app and accompanying care team–facing web interface, which was developed collaboratively by Takeda and American Academy of Family Physicians National Research Network (AAFP NRN) and powered by Fora Health (Ctrl Group Limited UK). The control arm (n=2 practices; 1 Community Health Center and 1 private practice) continued usual care for depression. Participating practices were recruited from the AAFP NRN. Patient interview participants were recruited from the patients enrolled in the intervention arm via email invitation in the final weeks of their study involvement. Practice champion clinicians, study coordinators, and clinicians who cared for patients enrolled in the study were invited to participate postintervention.

This study was approved by the American Academy of Family Physicians Institutional Review Board (Protocol 20‐389). All participants consented to be in the study. However, those participants who took part in the intervention were sent an email asking if they would be willing to participate in an interview. Those who responded received an invitation for a scheduled interview. Prior to the start of the interview, a study team member thoroughly described participation: expected interview length, questions asked, permission to record, and participant rights (voluntary participation, can terminate participation at any time without penalty, remuneration, and possible risks). Risks, such as discomfort answering certain questions and the potential loss of anonymity and confidentiality, were explained. Verbal consent was then obtained. On completion of the interview, participants received compensation; a US $100 e-gift card was sent to their email. All identifying information was removed from interviews and study data to protect participant anonymity.

The Fora Health app supports goal setting and tracks daily medication adherence and side effects, weekly goal attainment, and behavioral health participation; assesses the 9-item Patient Health Questionnaire (PHQ-9) [22,23], 13-item Patient Activation Measure [24,25], World Health Organization-Five Well-Being Index [26,27], and Perceived Deficits Questionnaire [28,29] on a biweekly basis; and provides short educational tutorials about depression management on topics such as talking with your physician and goal setting. Patient participants began using the Primary Care Path app (powered by Fora Health) after meeting with the study coordinator to enroll and set goals. The Primary Care Path web interface allows staff and clinicians to see patient-reported data. The expectation was that the care team would use the information before visits or review between visits.

Separate semistructured interview guides were developed by the AAFP NRN study team for patient and care team participants, according to the RE-AIM (reach, effectiveness, adoption, implementation, and maintenance) evaluation framework [30-32]. The interview guides were reviewed and revised by the broader study team. Interviews were conducted from May to October 2021 by a family physician (AN) and 2 research project managers (AML and EAR), each with a master’s degree and training and experience in qualitative methods. Interviews were conducted online and lasted approximately 30 minutes, and were audio recorded and transcribed. Four research team members (MKF, AN, AML, and EAR) met frequently over 13 weeks to develop the codebook, using a priori and emergent codes. Meaning, some codes were generated based on the discussion guide, and others were developed through open coding, where codes emerge from the data. After the research team members agreed on the codes and their definitions, each transcript was coded once (MKF and EAR) using Atlas.ti.9 (Berlin and Germany), and an audit was conducted (AML) to ensure intercoder reliability. This inductive approach to data analysis was completed by 4 members of the research team (MKF, AN, AML, and EAR), using a combination of grounded theory and constant comparative methods [33,34]. Themes emerged through repeated reading of the data. Themes were discussed and agreed upon by the broader research team.

The study team interviewed intervention patients (n=8) and intervention care team participants (lead clinicians, study coordinators, and other clinicians; n=8, shown in Tables 1 and 2). The sample size was determined based on the 2 participant categories of the sampling frame and the sample’s sufficiency to provide a meaningful perspective and nuanced detail [35]. Data comparisons between the patient and care team participant strata showed similarities and differences for topics of app/interface usability (navigation and content and usability), tracking (medication adherence, side effects, goals, suggestions for additional tracking features, and communication), and program recommendations. Saturation was reached within the patient stratum as there was consistency across all patient data. Even though much of the data indicates overall agreement, additional explanations are provided where discordance is present. Example quotations are displayed at the end of each section.

Patient and care team participants believed using the app could benefit several groups, namely those who are new to taking or switching medications, have trouble remembering to take medications, are proactive in their care, want to make a behavioral change, need reminders, need accountability, and are seeking increased adherence. Some care team participants recommended targeting patients who needed closer follow-up and thought the app was more beneficial for those who were more diligent about their depression care and management, as the increased engagement gave them better insight into their health. Some care team participants indicated that using the app interface did not impact or change setting goals, reaching goals, making shared decisions, or patient outcomes. Reasons provided were either that their practice had implemented a patient-centered approach previously or that they needed data to compare before and after study participation to make a true assessment.

Care team participants thought the workflow around the interface could be improved. They overwhelmingly stated the need for electronic health record (EHR) integration because logging into the interface interrupted workflow, making it impractical during a patient appointment. Some care team participants admitted they did not access patient participant information: logging in was not well integrated into their workflow, they did not remember to log in, or they did not know how to log in. For example, one care team participant in the residency practice commented that it would have been helpful for the attending physician to see the data, and one patient participant stated that they wanted to share data with a psychiatrist outside of their primary care practice. Furthermore, some care team participants spoke about the need for app support within the practice. One care team participant stated that the practice had a dedicated behavioral health care specialist who carried out nonmedication interventions, and much of the program fell under their purview (eg, providing information to physicians and following up with patients). It was stressed that behavioral health is a team effort, and office staff are needed for integration.

Overall, patient participants who tracked and care team participants who used data to inform conversations saw value in shared decision-making and setting goals and thought these tenets could be applied to other conditions, such as anxiety, weight management, diabetes, high blood pressure, attention deficit disorder or attention-deficit/hyperactivity disorder, and medication compliance.

Through this study, we identified barriers to and facilitators for implementation of a pilot program that incorporates a patient-facing mobile app and care team web interface to support depression treatment and management in primary care practices. These results on patient and care team participant perspectives of program implementation contribute to the rapidly expanding field of the use of digital health for depression care [17,36-45]. While some results between patient and care team participant strata were rather uniform, others differed. Our findings indicate that tracking medication adherence, side effects, and goals could allow both patient and care team participants greater insight into depression treatment and management. However, though the app may have helped some patient participants compile information for tracking, which may have led to more focused visits, not all care team participants thought app use improved patient outcomes. This may be more a matter of perception—perhaps patient participants who engaged regularly with the app thought they had better outcomes, possibly related to the perception of having more connectivity with their care team. Care team participants reported they could not definitively determine whether the app and program impacted patient outcomes, patient relationships, shared decision-making, goal setting, or goal attainment without tangible pre- and postintervention data. Nonetheless, medication adherence and goal setting are often related to depression outcomes and can be challenging [46-48]; this program could be another tool in their toolbox. Practices that already have patient-centered approaches to care may find aspects of this program duplicative; yet practices that do not prioritize approaches to goal setting and shared decision-making may experience more benefits if implementing this program, or a similar one.

App data in and of itself may not impact patient outcomes; it may be that the increased follow-up and touchpoints promote improvement in depression symptoms, as is evident in the literature [49,50]. The program was designed to connect patient participant data to the care team, allowing the care team to get real-time insight into patient symptoms and functioning between appointments. Thus, care team members who receive and review app data may be better able to recommend follow-up, suggest care changes, inform conversations, and shape next steps versus those who do not use additional data points collected outside of a care visit. The novel contribution of this study is that care team members who viewed near real-time information had an opportunity to make more timely decisions, especially for medication changes. Therefore, data insights provided to the care team via the interface may allow treatment pivots based on more accurate data on what is or is not working. This finding suggests that those care team members who accessed patient information in near real-time could make more prompt evidence-based decisions, which is particularly important for a patient with depression symptoms that impair day-to-day functioning.

Even though care team participants appreciated aspects of the program, they agreed that successful implementation would need to be more integrated into existing workflows. It is unclear to what extent care team participants, especially physicians, accessed data prior to and during visits, and they often said that accessing a separate website was too time-consuming during a busy clinical day. This is consistent with other literature that demonstrates that app data integration into clinical workflows is often a requirement for use due to heavy workloads [51-55]. EHR integration would allow all care team members access, a seamless transition for viewing data, and a comprehensive view of patient data. For example, if a patient scheduled an appointment for something unrelated to depression, a care team member would not necessarily review their depression data on a separate website. It is well documented that EHRs can help clinicians prepare for appointments, making them more efficient [56,57].

For uptake of this program, other workflow processes may need to be established depending on the practice, such as the level of behavioral health integration or resource support (eg, access to behavioral health professionals or referrals to community resources) [58-61]. Another consideration is designing the best approach for using data in conversations with patients [17]. Future iterations of this program could include additional education for care team members regarding roles, recommendations, and patient interactions.

This study has limitations. First, our study sample was small and was racially and ethnically homogeneous, limiting the generalizability of the findings. Second, data saturation was reached in the patient stratum but not the care team member stratum. The care team contained diverse roles. We interviewed 8 primary care physicians and clinicians; their experiences varied. For example, a behavioral health professional or care coordinator may have had a touchpoint with the patient, and physicians may or may not have used the interface. Therefore, it is inconclusive as to whether physicians accessed and used the data before or during visits. To help remedy this situation, care team participants recommended embedding the interface data into the EHR. Third, to some extent, these data were influenced by patient and care team member engagement with the program. For example, if a patient used the app for a limited time span, they may not have tracked their behaviors long enough to see trends or for care team participants to make recommendations. Finally, a patient’s depression cycle may have affected app use and program involvement. If depression symptoms were more prominent, patient participants may not have had the motivation to track medication adherence, side effects, or goals.

Implementation of a digital health program for depression treatment and management in primary care practices may help support patient medication adherence, facilitate timely medication changes, encourage goal setting, and foster communication between patients and care team members. While patient and care team participants valued program tenets, enhancements such as minimizing workflow interruptions, integrating data into the EHR, providing education and best practices for patient interactions, augmenting content with helpful resources, and adding personalized content that evolves alongside patient progression could increase program engagement.