This study was approved by the American Academy of Family Physicians Institutional Review Board (protocol 20‐389) under categories 5 and 7. Participants were enrolled and consented electronically via a waiver of documentation of consent process. Participant confidentiality was protected by storage of data within a secure information technology environment and maintained as participant identifiers were removed for analysis. Patients received e-gift cards in the amounts of US $30 for completing the enrollment survey, US $25 for each of the 4 follow-up surveys, and were given a US $40 bonus if all 4 follow-up surveys were completed on time. E-gift cards were sent via an automated email after the completion of the survey from RewardsGenius/Tango Card.

Intervention practices used the Primary Care Path program, which consisted of a patient-facing mobile app and accompanying care team website (subsequently referred to as “the program”) with their participants to facilitate SDM, symptom tracking, and medication adherence. Practices in the control group did not use the program but recruited patients to respond to study surveys over an 18-week period. The app allowed clinician entry of goals created by patients in conversation with their clinician. Patient participants were able to track medication adherence (daily) and mental and physical symptoms on a daily or weekly basis. Goal setting was undertaken with clinical care team members, and goal attainment was tracked on a weekly basis within the app. Clinical care team members at intervention practices were able to see up-to-date information from patient app use when they accessed the program.

Descriptive statistics (counts, percentages, means, and standard deviations, as appropriate) were calculated for all questions and scores. Exploratory analyses consisted of chi-square tests (dichotomous variables), ANOVAs (categorical variables more than 2 categories), and ANCOVAs (controlling for demographic variables) were used for inferential statistics to determine single time point results using the full data at each time point (group results; separate analyses were performed for High Engagers as well as for all Intervention against Control). For additional exploratory analyses, mixed ANOVAs were used for analyses over time and used data at the individual level; an individual had to have data at each time point to be included in the analysis. Mixed ANOVAs, which require all data to be present, were performed for the following combinations: (1) baseline to 6 weeks, (2) baseline to 12 weeks, (3) baseline to 18 weeks, (4) baseline to 6-12 weeks, and (5) baseline to 6 weeks to 12 weeks to 18 weeks. Missing data were removed casewise for each analysis. Tukey HSD and Games-Howell were used for post hoc analyses. For all analyses, all data collected were included. However, for the tests that were tracking a person over time (ie, mixed ANOVA), the respondent had to complete each assessment included in the analysis. The mixed ANOVA was a per-protocol analysis. While these were exploratory analyses due to the pilot or feasibility nature of the study, we did apply the Bonferroni correction for each comparison and noted where it did and did not affect results. An α value of .05 for 2-sided tests was used for statistical significance. SPSS (version 27; IBM Corp) was used for all analyses.

Intervention patients used the app with a wide range of engagement. Interaction touchpoints ranged from 0 to 126 throughout the course of a participant’s involvement in the study. App use data yielded 2 levels of app engagement among intervention participants. A natural cutoff in the data was found with <7 app touchpoints corresponding to low engagement and ≥7 touchpoints indicating high engagement (“High Engagers”). This natural cutoff also corresponded to intervention patients interacting with the app for less than 2 weeks (data not shown). The High Engagers group was established as a post hoc stratification once it became clear that there were 2 distinct levels of usage among the intervention patients.

The majority of the participants recruited were female (Control: 32/42, 76.2%; All Intervention: 27/34, 79.4%; and High Engagers: 22/27, 81.5%) and White (Control: 40/42, 95.2%; All Intervention: 31/34, 91.2%; and High Engagers: 22/27, 92.6%; Table 1). Participants were equally distributed in relation to gender and race across groups (gender: Control/All Intervention: P=.95, Control/High Engagers: P≥.99; race: Control/All Intervention: P=.43, Control/High Engagers: P=.70). However, there were Hispanic/Latinx participants in the All Intervention/High Engagers group but none in the Control group (Control: 0.0%, 0/42; All Intervention: 14.7%, 5/34; High Engagers: 18.5%, 5/27; Control/All Intervention: P=.01, Control/High Engagers: P=.003). Other demographic variables (eg, highest degree, marital status, type of health insurance, and income) did not differ between groups and are reported in Table 1.

No matter the time elapsed, all participants saw significant increases in feeling successful in achieving their treatment goals for depression. Between baseline and 18 weeks, the High Engagers group had a significantly larger increase in feeling successful than the Control group (Time: F_{1, 45}=14.072, P<.001; Time × Group: F_{1, 45}=4.101, P=.049). When including all time points over the study period, the High Engagers and All Intervention groups had significantly larger increases in feeling successful than the Control group (Control/All Intervention: Time: F_{3, 123}=7.436, P<.001, Time × Group: F_{3, 123}=2.70, P=.048; Control/High Engagers: Time: F_{3, 117}=7.236, P<.001, Time × Group: F_{3, 117}=2.698, P=.049). Between baseline and 18 weeks, Control group versus All Intervention group (Time × Group: F_{1, 55}=5.721; P=.02) and Control group versus High Engagers group (Time × Group: F_{1, 54}=6.029; P=.02) had significant changes over time in participants having discussions with their medical provider about their depression treatment goals with Control Group increasing over time and All Intervention and High Engagers decreasing over time (Table 3).

At 18 weeks, more participants in the Control group (11/32, 34%) indicated that they set treatment goals for their depression in the last 18 weeks than those in the All Intervention (4/24, 14%; χ²₁=3.2 [n=60]; P=.07) and High Engagers (2/22, 9%; χ²₁=4.6 [n=54]; P=.03; Table 3) groups. More participants in the Control group (9/24, 27%) had also discussed their depression treatment goals with their medical provider within the last 18 weeks than those in the All Intervention (2/28, 7%; χ²₁=4.2 [n=61]; P=.04) and High Engagers (1/22, 5%; χ²₁=4.6 [n=55]; P=.03; Table 3) groups. At all time points compared with baseline, all participant groups set fewer goals for depression with no significant difference between groups (Table 3).

While we report on a small-scale pilot study that is underpowered for conclusive and generalizable conclusions, the data presented here can inform hypotheses for future, fully powered studies to examine the integration of digital resources into primary care practice. We evaluated a primary care-centric program designed to promote the use of digital tools and resources to promote patient or clinician communication and support SDM, goal setting, symptom monitoring, and other aspects of patient self-management. We found that some results held after controlling for multiple comparisons, while others did not and prompt the need for further exploration. Differences that held include that, compared with the control group, coordinated use of the program by both patients and clinical teams in primary care practices was associated with earlier communication between patients and medical providers and increased feelings of success in achieving goals. Significant differences at the individual analysis level that did not hold after correction included that intervention patient participants received earlier medication changes and reported starting a medication for depression, changing medications, or changing the dose of medication more frequently. For both intervention and control groups, reports of participating in SDM did not differ and were highest at baseline and decreased over time. Finally, improvement of depression symptoms was the same between the intervention and control groups. These findings indicate that there may be some benefits of the use of the app itself (eg, feelings of success in goal attainment) and the program overall (app plus clinical use of the program; eg, communication and treatment changes). However, this study was a pilot and was not adequately powered to be able to determine whether or not these differences would be generalizable.

Despite recommendations and guidelines for its use, adoption of SDM is generally low in clinical care [19]. Here, we explored whether the coordinated use of the Primary Care Path program by patients and clinicians would facilitate and increase SDM in the intervention arm. We did not see a difference in reported quality of SDM between arms, but earlier communication between patients and clinicians and earlier treatment changes, potentially as a result of SDM, were noted in the intervention arm. While further exploration is needed, interviews with clinicians and patients in intervention practices indicated that they were using the program to inform treatment and management strategies [20]. SDM-Q-9 responses may not have shown differences for a few reasons: the program had no impact on SDM; the small scale of this study; the fact that both the control and intervention practices often participate in research projects, priming clinicians at all sites to be invested in quality improvement and implementation of SDM outside of their involvement in a given initiative or study [21]; and masking of program effects by confounding factors (eg, a multicenter study found that clinicians involved in committee work outside of medical practice do more SDM) [22]. More exploration of how or whether the program was used for SDM to support MDD management is warranted.

Interestingly, despite no difference in SDM being reported, we found that both All Intervention and High Engager groups reported decreasing conversations with their clinical teams over time, while Control patient participants reported increasing conversations over the study period. This held when correcting for multiple comparisons and suggests that All Intervention and High Engager groups may have addressed patient or provider communication needs earlier, while those without access to the Primary Care Path took longer to initiate or complete communication. These results suggest that the overall program may have facilitated or encouraged communication between the patient participants and their care teams, as we cannot tell whether the communication was initiated earlier by the intervention practice teams, the patients themselves, or both. It is noteworthy that conversations were occurring in both arms, but these conversations were not correlated to SDM, which decreased for all, as reported by SDM-Q-9 throughout the study.

Despite the finding that goal setting increased more over time in the control group, analyses showed differences in rates of setting goals at 18 weeks, suggesting that using the app may have supported goal setting. This could be a feature of using any app aimed at aiding in self-management, although the Primary Care Path app specifically featured goal setting and provided a format to set and track goals. When patients have a way to record and track their goals, they may be more cognizant of their progress on those goals, thus explaining the finding of more self-reported success with goal attainment among the intervention group. Further exploration in a fully powered study is needed to better understand the role of the app and program on goal setting and attainment. However, our finding that both High Engagers and All Intervention participants reported feeling more successful in attaining their goals suggests that the coordinated use of the program by practices and their patients supported participants in recognizing their progress toward goals throughout their involvement in the study.

Treatment changes (starting, switching, or decreasing the dose of an antidepressant) were more likely for participants in the intervention by All Intervention and High Engagers analyses at multiple time points, suggesting that the intervention may have facilitated medication management for depression treatment. Results indicating that communication may have been more frequent earlier for patients in the intervention arm may have played a role in earlier medication management. There is also evidence that symptom tracking by itself can positively influence health outcomes by increasing patient self-efficacy [23], and that patients who track their symptoms regularly and share their data with their clinician can feel that they receive more personalized, thus effective, care [24]. However, there is a lack of evidence on the effect of patient symptom tracking on treatment changes. It is also possible that recording symptoms and medication use made patients more aware of (1) day-to-day depression symptoms that might warrant a trial of medication and (2) side effects or other factors that might influence considering a change of medication. While communication differences were reported between arms, there was no difference in appointment frequency, so it seems likely that medication changes were made for specific reasons that were discussed during or between appointments.

It is unclear whether the positive outcomes related to symptom tracking that have been reported by similar studies are due to unreported treatment changes as a result of patient-recording data, or whether the data here are unique due to the features of the Primary Care Path program itself. The continuous nature of the symptom tracking in the app and the accessibility of this information to the clinician may have also facilitated earlier treatment changes by providing clinicians more evidence to necessitate treatment change. To our knowledge, at the time of this study, this was the only app focused on the management of MDD in primary care that automatically shares patient-reported data directly with their care team; however, further research is warranted to verify the effect and causality.

We acknowledge several study limitations. First, this is a pilot study and by nature includes nonrandomized practices and a small sample size, thus limiting the statistical power and increasing the potential for selection bias. Selection bias is possible with the small number of practices and sample size here. The goal of the study was to pilot-test the program for future use and garner data necessary to fully power a larger study where we can reduce selection bias from multiple fronts. The need for participants to own and be able to use a smartphone also affects the generalizability of our findings; however, smartphone ownership was at or near 85% during the course of this study, with rates above 60% for both low-income and older adults [25,26]. Another limitation is that the study design did not require clinician visits, so the use of or reference to patient-entered app content may have occurred in or outside the context of a clinic visit, thus impacting responses to the SDM-Q-9. Workflow changes were required for intervention practices, which may have introduced unknown factors that influenced the way the program was used. Examining variations in app use by patients may have been better addressed a priori to reduce the risk of type I error and reduce the potential for effects such as generalizability and error rate of the statistical tests. One of the most noteworthy limitations of the study is that it was initiated in the early phases of the COVID-19 pandemic, leading to circumstances that affected health care delivery and study implementation. Practice recruitment, staff engagement, and patient willingness to participate in a study may all have been affected by the implications of living through and providing medical care during the pandemic, although the pandemic may also have facilitated the adoption of the program and use of telehealth for this study. While these are problems that are commonly encountered in real-world studies, they also impact results, validity, and replicability. Limitations notwithstanding, this pilot study provided learning opportunities and findings that can be comprehensively examined in a fully powered study.

This program, which facilitates clinician access to patient-generated goals and information, may enhance MDD management in primary care, leading to earlier medication optimization and patients feeling that they have achieved the treatment goals that matter most to them. This study contributes to beginning to understand the nature of coordinated use of digital tools by clinical practice teams and patients to improve outcomes for MDD in primary care. To clarify some conflicting findings and verify and determine the generalizability of results, a fully powered study with a larger sample of participants and practices is needed.