Telemental Health Collaborative Care Medication Management: Implementation and Outcomes

Self-Reported Measures

Patients completed electronically secure online assessment questionnaires at baseline and between follow-up visits. Assessments consisted of the PHQ-9 and the GAD-7, well-validated measures of depression and anxiety.^29,30 For depression, scores of 4 and below indicate no to minimal symptoms, 5 to 9 indicate mild symptoms, and above 10 are moderate and severe symptoms. For anxiety, the score of 4 and below on the GAD-7 indicate no to minimal symptoms as well; a clinical cutoff of 8 was used for the GAD-7 for moderate to severe symptoms, as research suggests that scores at or above 8 are highly likely to correspond to an anxiety disorder diagnosis.^31,32

Treatment

The MM program combined live video-based sessions with Board Certified Internal Medicine and Family Medicine physicians plus technology-based care education. These included digital video lessons (e.g., on understanding anxiety) and exercises/guides (such as self-assessing for goals), as well as assessments/check-ins (such as symptom reporting) that clients had access to use in between sessions; physicians were trained in assigning digital content to supplement clinical care. All treating physicians had access to a psychiatrist consultant for case and care review. All physicians were provided with continuing medical education before accepting clients focused on evidence-based practices and guidelines. Physicians had protected time for administrative duties, completion of thorough notes, review of patient data between visits, and care coordination with other providers.

In traditional CoCM models, a 30- to 60-min intake and shorter follow-up would address all health concerns, in addition to mental health; this model offered a deep dive into mental health with a 90-min intake and 30-min follow-ups. Another aspect novel to this model is continuous quality assurance, training, and consultation: sessions were recorded with participants’ consent, and quality adherence was ensured through session review, weekly supervision video calls, and video consultation groups with other physicians.

Digital Components of MM

MM sessions were conducted on a secure, Health Insurance Portability and Accountability Act-compliant video platform. Following each session, physicians assigned information sheets about medications and assessments to be completed between sessions. They could also provide digital lesson videos, which consisted of animated videos and quizzes to test for comprehension and provide corrective feedback. The digital lessons utilized a storytelling approach where viewers followed the therapy journey of multiple characters presenting with symptoms of depression or anxiety. This approach has been used in other efficacious internet-delivered cognitive behavior therapy interventions, and it was found to have a normalizing effect for clients.³³

Data Analyses

To estimate the impact of MM, we conducted paired two-tailed t-tests between baseline and last available assessment scores for each measure. We calculated Cohen’s d_rm, a conservative measure of effect size for within-subjects designs that controls for the correlation between measurements. We also summarized the response (50% reduction in PHQ-9 or GAD-7) scores at follow-up and remission among those who had measurable anxiety or depression at baseline and were subthreshold at follow-up (<5 on PHQ-9 or GAD-7).

Results

Two hundred twelve participants completed the initial consultation, and had a follow-up PHQ-9 and GAD-7 assessment at least 8 weeks from their initial assessment. Participants’ mean age was 32.00 years (standard deviation [SD] = 7.38); 58.96% (n = 125) of participants were women and 41.04% (n = 87) were men. Average number of visits was 5.74 (SD = 2.54). Around 58.02% of clients had at least two therapy sessions between the first and last assessments included in the analysis and 0% of clients had any coaching session in that time (Table 1).

Depressive Symptoms

At pre-treatment, mean PHQ-9 score was 13.02 (SD = 5.37). Approximately 26.42% of participants reported some amount of suicidality on PHQ-9. Breakdown of depression severity according to PHQ-9 was 25.94% with mild/minimal depression (PHQ-9 < 10), 74.06% with moderate/severe depression. Results of paired t-tests revealed that for people with moderate-severe depression, PHQ-9 scores decreased from baseline to follow-up assessment. At 12 weeks, participants improved an average of 7.27 points (SD = 4.80), t(145) = 18.30, p < 0.001, Cohen’s d_rm = −1.64; and at 24 weeks, scores improved an average 7.17 points (SD = 5.00), t(65) = 11.64, p < 0.001, Cohen’s d_rm = −1.54, both suggesting a large effect of treatment on depression symptoms (Table 2). Of people with moderate-severe depression, 58.90% had a positive response at 12 weeks and 51.52% had a positive response at 24 weeks; 24.66% had remission at 12 weeks and 24.24% had remission at 24 weeks (Table 3).

Anxiety Symptoms

At pre-treatment, GAD-7 score was 11.73 (SD = 5.28). Breakdown of anxiety severity according to GAD-7 was 24.06% with mild/minimal anxiety (GAD-7 < 8), 75.94% with moderate/severe anxiety. Results of paired t-tests revealed that for people with moderate-severe anxiety, GAD-7 scores decreased. At 12 weeks, participants improved an average of 6.71 points (SD = 5.18), t(148) = 15.80, p < 0.001, Cohen’s d_rm = −1.51; and at 24 weeks, participants showed an average improvement of 6.03 points (SD = 5.37), t(69) = 9.39, p < 0.001, Cohen’s d_rm = −1.35, both suggesting a large effect of treatment on anxiety symptoms. Of people with moderate/severe anxiety, 51.68% had a positive response at 12 weeks and 40.00% had a positive response at 24 weeks; 29.53% had remission at 12 weeks and 32.86% had remission at 24 weeks.

Response and Remission

Of people with either moderate/severe depression or anxiety at baseline, 65.88% of patients demonstrated response either on PHQ-9 or GAD-7 at 12 weeks and 56.41% at 24 weeks; 44.71% demonstrated remission either on GAD-7 or PHQ-9 at 12 weeks and 41.03% at 24 weeks. Around 68.24% and 56.41% experienced response or remission on either scale at 12 and 24 weeks, respectively.

Discussion

In a sample of 212 patients treated in a collaborative and technology-driven model, responses to mental health MM were appreciably higher than traditional modalities of care. Prior medication trials have reported response rates of 41–47%, and remission rates of 13–35% at 3 months of medication use for depression.^34,35 For anxiety, prior literature has reported response rates of 63% and remission rates of 35% at 3 months of medication use.^36,37 (Of note, the specific outcome measures in those studies may be different from the GAD-7 and PHQ-9.)

In research CoCM, response and remission from depression have been reported to be 35–46% and 20%, respectively.^7,38 In more representative (not study selective) settings, the response and remission rates have been reported to be as low as 19% and 5%.³⁹ In contrast, the results presented in this study achieve higher rates of response (>50%) and remission (>25%) in people with depression or anxiety at 12 weeks.

This is the first study examining outcomes in a collaborative and technology-driven mental health medication prescribing model that integrates digital tools, digital psychoeducational lessons, and measurement-based care and is focused only on mental health. As noted in the introduction, barriers to CoCMs include provider dissatisfaction, difficulty tracking outcomes, lack of communication, lack of education, turnover, and insufficient staff time, especially when addressing mental health, in addition to other health concerns.^7–10

The model presented in this article differs from traditional CoCM models. While we do integrate elements of team, population, evidence, and measurement-based care, we focus the visit on mental health, rather than physical and mental health.¹³ The physicians in this model are Family Medicine and Internal Medicine physicians, so they have a broad understanding of medicine and complete a full medical history, but the visit focuses on the mental health concerns.

While in traditional models, a 30-min intake and 15-min follow-up would address all health concerns, in addition to mental health, this model offered a deep dive into mental health with a 90-min intake and 30-min follow-ups as clinically indicated. As a result, the physician can fully integrate guidelines, including a full review of psychiatric symptoms, history, substance use, risk assessment/planning, medical items, measurement-based care, shared decision making, and effective documentation.¹⁷

Time is uniquely protected for case consultation both individually and in a group with the psychiatrist, as well as for coordination of care with other providers (therapists, primary care physicians, etc.). (There was no routine case manager involved in the care.) The digital platform allowed for measurement-based care and seamless communication. All of these adjustments to traditional CoCM models may have improved outcomes. With only 1,907 psychiatry resident position offered in the 2021 National Resident Matching Program Main Residency Match versus 13,847 Internal or Family Medicine residency positions, we can meet the mental health shortage with creative solutions, combining psychiatrist leaders and mental health passionate primary care champions.⁴⁰

Barriers to implementation center largely around adoption of services. Having members come to care continues to be a challenge; while we know that a greater percent of eligible members may benefit from a medication consultation by virtue of their severity, only 0.16% sought a consultation. We have instituted several efforts to increase member and provider (therapist) awareness to increase adoption. On the physician side, consulting with physicians in real time (during a visit) while we are at different time zones can be challenging, but we use messaging systems to aid; we also have established individual and group consultation times that are protected for case review.

Clinically, the other challenge has been with individuals who are interested in controlled substances, which we do not provide in the platform due to the Ryan Haight Act; those individuals are referred to our in-person network of providers. Feedback from members has been positive with the majority rating their visit as a 5 on a 5-point scale, and within our physicians’ group, we have had 50% of the physicians from these data transition to full time from part time.

Aside from inherently being an uncontrolled data set, which limits the interpretation of the findings, another major limitation of the analysis is the inconsistency of follow-up periods. Among those with follow-up assessments, the timing of the end point is variable. To be conservative, we assumed an individual’s last assessment was the endpoint, even if they remained in care for far longer. That endpoint may be well before the last session or up to 5 weeks after the last session. Therefore, comments on outcomes are reliant upon this assumption of timing. To add, if an individual’s scores vacillated in treatment, that is not captured in this analysis.

It is also notable that improvement generally dropped between 3 and 6 months, possibly due to either worsening outcomes or response bias, where patients who are improved may be less likely to complete outcome assessments. Another limitation is about the chronicity of care. Some people seeking MM had plateaued with therapy, whereas others did not engage with therapy ever or engaged later. Changes in scores do not account for what part is attributable to MM only. Given that 58.02% are concurrently in therapy during their episode of care with MM, it is reasonable to assume that some of the gains in the outcomes were related to therapy, or that there was a synergistic effect from concurrent pharmacotherapy and therapy.⁴¹

The therapy service availability is a positive aspect of the model and another distinguishing feature of this analysis is that we have access to detailed data about therapy usage, which may not otherwise be available in the community. Future studies could compare similar cohorts in MM, therapy, and both to comment on attributable gains.

This study focuses on depression and anxiety outcomes. Other common diagnoses among those who had measurable anxiety or depression at baselines were bipolar (4.52%) and attention deficit hyperactivity disorder (13.87%). By limiting outcomes to PHQ-9 and GAD-7, we do not account for variability in outcomes for diagnosis-specific conditions, and instead offer a simple model.

In terms of population, the included sample may be unique. Unlike the preponderance of care models, MM participants have easy access to therapy in all cases (>50% of client searching for therapy in the care platform will be offered as appointment in the next 2 days with an average wait time of 6 days between booking and the first completed therapy appointment), so they may elect to complete a medication consultation solely due to personal preference. Some may be in therapy or may enter therapy. We noted that 13% of American adults take antidepressants, but 60% of Americans with mental health issues do not get the treatment that they need^3,4; the choice of utilizing a psychiatric medication in the community may be in the absence of other resources and comes from a primary care physician without protected time or support.

Therefore our results may be impacted in one of two ways. First, included clients may self-select as more severe and want to access medications, while in usual care settings, medications may be the first point of access without therapy. On the other hand, these clients may have more robust outcomes because they have access to therapy resources.

Future research would account for covariables, which may affect outcomes such as therapy concurrency, choice of medication, diagnosis, completion of digital psychoeducation materials, impact of direct messaging with providers, and frequency of visits. A prospective randomized design would shed more light on the efficacy of this model. If limited to a cohort design, it would be useful to summarize data of a larger group over a longer period of time. Of note, that while 269,794 people were eligible for care in our model, only 433 (0.16%) sought a medication consultation. Since this service is efficacious and convenient, dissemination through awareness campaigns may help people connect with mental health services sooner.

Conclusions

To summarize, developing a technology-enhanced CoCM service available to employees/dependents through their employer was feasible and demonstrated favorable outcomes on depression and anxiety. This type of digital clinic with access to therapy and coaching can improve access and reach of mental health. This model allows for better implementation of evidence-based care, designated time for collaboration, and preservation of the therapeutic relationship.

Authors’ Contributions

S.D. contributed to the research design and implementation, data acquisition and analysis, and article drafting and revisions. J.W. contributed to the article drafting and revisions. S.-Y.C. contributed to the data acquisition and analysis. C.C. contributed to the research design and implementation, data acquisition and analysis, and article drafting and revisions.

Disclosure Statement

S.D., S.-Y.C., and C.C. have been granted equity in Lyra Health. J.W. has no ongoing financial relationships to disclose.

Funding Information

This study was funded and executed by Lyra Health, Inc.