The public health burden of depression and anxiety is enormous and rising, with effects extending to unemployment, risky substance use, and of greatest concern, suicide. In particular, young adults have seen a substantial increase in anxiety, depression, and suicidal thoughts or behaviors over the past decade [1,2]. Mental health care systems, globally, face significant challenges in meeting the staggering need for care created by this rising public health burden. In particular, access to traditional forms of mental health care is limited by cost, time, effort, and stigma [3], and the number of clinicians trained in evidence-based treatments is grossly inadequate to meet current needs. The urgency of these needs has stimulated the development of 2 broad strategies for increasing access to care: stepped care approaches and digital provision of therapies.

In stepped care models, typically, all patients begin with low-intensity treatments and only nonresponders are moved to the next (more intensive) step of care [4]. Low-intensity treatments often include digital or online therapies, with or without support, as multiple trials targeting anxiety and depressive disorders have shown that these treatments can be as effective as face-to-face therapy [5-8]. While this “fail-first” approach more efficiently allocates resources than traditional approaches that strive to provide the highest level of clinician-delivered care to all patients, it results in delays to more intensive treatment for those who need it [9], and such delays may lead to symptom worsening and increased disability [10], discouragement for new treatments after initial options fail [11], or dropout before a second course of treatment [12]. An emerging body of literature indicates that stratified models, in which individuals are matched to level of care based on predetermined criteria (typically current symptom severity), produce superior outcomes and are more cost-effective in treating depression and anxiety compared with fail-first stepped care models [13-16].

The use of data collected from patients during the course of treatment for clinical decision-making (measurement-based care) has the potential to greatly improve outcomes but remains the exception rather than the rule [17]. Specifically, measurement-based care typically outperforms usual care on both symptom reduction and reduced dropout rate [18] and is associated with decreased costs and lower odds of patient deterioration during treatment [19]. Without measurement-based care, clinicians have difficulty identifying patients who are at risk for nonresponse or deterioration [20,21]. Consequently, interventions become reactive to “crisis” needs rather than being proactive. Further, decisions about when treatment has been successfully completed tend to be determined by when the patient or their provider “feels ready” to terminate therapy, which may lead to extending treatment far beyond the point when symptom gains have been attained and maintained.

Even when measurement-based care is utilized, guidelines on how to adjust treatment over time are lacking. As discussed, stepped care approaches typically personalize level of care after a patient has shown nonresponse. Waiting several months to make adaptation decisions could have dire consequences for individuals with depression and anxiety [22]. Moreover, typical stepped care approaches step up only for patients who remained actively engaged in treatment, thereby not addressing the critical issue of how to manage and engage individuals who prematurely discontinue treatment [23].

Furthermore, typical care models fail to fully consider changes in clinical status after the end of acute treatment; relapse is not uncommon (especially for depression), yet rarely are patients monitored in order to identify for whom reinitiation of care is warranted. By identifying those at risk for nonresponse or worsening of symptoms during or after treatment, adaptations can be made to increase treatment intensity or reenter treatment and prevent further worsening of symptoms and their potential consequences. Routine monitoring of clinical status enables dynamic adaptation of treatment as needed, which has the potential to improve effectiveness and reduce attrition, as patients may be more engaged in treatment when they are receiving what they need most at the time they most need it [24-26]. Adaptive interventions can also enhance cost-effectiveness by increasing the efficiency of service delivery and reducing downstream service costs [27].

Suicidal thoughts and behaviors are a major public health problem, particularly among young adults [28]. Rates of suicide attempts, nonsuicidal self-injury, and suicidal ideation increase markedly during adolescence [29-31], with suicide death rates increasing as youth move into adulthood. Across all adult age groups, the prevalence of serious suicidal thoughts is highest among young adults aged 18-25 (11%) [32]. There is a critical need to embed methods for detection of risk and strategies for suicide/self-harm prevention within systems of care, which go beyond relying on patients to reach out to providers or to call crisis lines, and beyond monitoring at frequencies that fail to capture tipping points toward high risk. Digital tools may be particularly useful in assessing and monitoring such risk, given that they can promote increased self-disclosure of sensitive topics compared with face-to-face assessment and allow for more rapid, in-the-moment identification and response [33].

Screening and Treatment for Anxiety and Depression (STAND) is a stratified stepped care model that incorporates online screening, continuous symptom monitoring over ~10 months, and tiered treatment for anxiety and depressive symptoms, with ongoing suicide risk detection and management (Figure 1). After individuals complete a brief online adaptive assessment of symptom severity and suicide risk (selected modules from the Computer Adaptive Test for Mental Health [CAT-MH]; [34]), they are provided feedback, routed to an appropriate level of care, and scheduled for an orientation or intake session, all within minutes of completion of initial screening. The CAT-MH assessment continuously tracks symptoms (both during and after active treatment) to adapt treatment (ie, move to a higher level of care when symptoms worsened or to a lower level of care for maintenance and relapse prevention) and to rapidly detect and respond to signs of elevated suicide risk. For the open trial described herein, there were 4 levels of care, or tiers.

Tier 0 was for those with no or minimal symptoms of depression or anxiety on the initial CAT-MH assessment. These participants were offered the option to continue completing the CAT-MH biweekly [34]. Tier 1 was offered to those with mild depression or mild anxiety on the CAT-MH, and included digital therapy. Coaching was offered to tier 1 participants given evidence for its positive effects on retention and clinical outcomes from digital therapy [8]. Tier 2 involved digital cognitive behavioral therapy (CBT) with advanced (doctoral student) coaches, and was offered to individuals with moderate depressive symptoms or moderate to severe anxiety symptoms. Evidence-based psychological treatment with option for medication management was offered to those with severe depressive (or manic) symptoms or suicidality (tier 3). Core active ingredients of CBT (see the “Methods” section) were selected to match each participant’s problems areas and were delivered by clinical psychology graduate students. Therapy was complemented by protocolized medication management as needed (modified from the STAR*D trial; [35]), delivered by psychiatry residents.

Across all tiers, treatment was adapted according to weekly or biweekly CAT-MH scores (Figure 2). If participants’ scores indicated a need for a higher tier of care at any time, they were contacted to initiate switching to the appropriate level of care. Similarly, participants in tier 3 could be switched to a lower tier and were thus offered access to digital therapy materials when they completed a course of tier 3 treatment and symptoms had shown improvement. Moreover, within tier 3, clinicians used weekly CAT-MH scores to guide decisions regarding treatment strategy (such as switching from behavioral activation to cognitive restructuring).

Frequent monitoring of suicide risk was conducted using a standardized protocol for responding to risk in real time (see the “Measures” section for details). This approach to risk management was implemented for all levels of care within STAND. In addition to outreach, a positive suicidal triggering alert indicated adaptation to tier 3 if the participant was in a lower tier.

STAND improves standard clinical practice paradigms in several ways. First, STAND provides increased access to evidence-based care by utilizing digital tools for a large proportion of the patient experience from screening through service enrollment to service delivery and risk management. Second, the emphasis on adapting both the level of care and intervention type as needs change is a major improvement over existing approaches (eg, stepped care), which typically wait until nonresponse is established at predefined points that are months apart before offering an alternative, if at all [16]. Third, STAND focuses on maintaining gains and minimizing relapse, which are rare in systems of care, despite the many benefits to individuals and a health system [36]. Finally, the STAND program focuses on personalization of services based on continuous symptom monitoring and clinical decision-making (both with regard to tier triaging and adaptation as well as choice of treatment modality within tier 3, the highest level of care) that is expected to reduce attrition and improve symptomatic and functional outcomes [23].

This study aimed to develop and evaluate the preliminary acceptability, feasibility, and effectiveness of the STAND program in a large sample of college students in an open trial. It was hypothesized that participants would report high satisfaction with the program, and would show significant improvement in anxiety, depression, and suicide risk severity scores over the course of participation in STAND.

Recruitment and enrollment into the open trial occurred between Fall 2017 and Spring 2020. All University of California, Los Angeles (UCLA) students were made aware of the screener through announcements from UCLA’s Chancellor, tabling at campus events, campus residence hall flyers, classroom announcements, web-based advertisements, banners, and email blasts. Students completed the CAT-MH online and received immediate feedback regarding their scores and recommended tier of care. Those who were interested in and eligible for the study and who provided informed consent were then scheduled for a baseline visit. At the baseline visit, research staff gathered pertinent information about treatment history and other demographics, after which participants were notified about how to start care in their assigned tier, with ongoing completion of the CAT-MH and other assessments over the course of 40 weeks.

All matriculating UCLA students who were at least 18 years of age and had access to the internet were eligible to complete the initial screener. Additional inclusion criteria for screening were English fluency, willingness to install an app on their phone, and no plans (eg, upcoming long vacations or studying abroad) that would interfere with participation. Additional exclusion criteria following screening for those entering clinical care (ie, tier 3) were severe psychopathology requiring intensive day-treatment or inpatient care (eg, severe eating disorder or substance use disorder, multiple [2+] suicide attempts leading to hospitalization within the past 6 months, significant psychotic symptoms not part of a severe major depressive episode, or bipolar manic episode) and were determined at the initial assessment through further screening. Participants in tier 3 were also excluded if they were unwilling to provide a blood sample (for secondary analyses outside of the scope of this manuscript) or had a current psychiatrist or provider and were unwilling to transfer their care to the study team for the period that they will receive treatment in STAND.

Descriptive statistics were used to characterize the data and to identify indices of feasibility and acceptability. With regard to the primary outcome data, a series of hierarchical linear models (1 per tier per outcome measure) were conducted with time as the level 1 predictor and the outcome of interest (CAT-ANX, CAT-DI, and CAT-SS scores for tier 3, and CAT-ANX and CAT-DI scores for tiers 1 and 2). In line with an effectiveness paradigm, we included all participants who were assigned to a tier and had at least one CAT-MH datapoint, regardless of how many sessions/lessons were completed or how much missing data on the CAT-MH was observed. Parallel sets of analyses were run with and without a covariate (number of sessions/lessons completed on level 2). The survey data were not consistently collected for participants in tiers 1 and 2 after completion of the ThisWayUp program. Therefore, results for tiers 1 and 2 reflect CAT-MH scores over the course of ThisWayUp program completion only (ie, 8 weeks). Further, of the 11 participants who completed a baseline assessment for tier 0, only 5 participants completed more than 3 additional assessments. For this reason, CAT-MH scores are not reported for tier 0.

As a secondary analysis, the percentages of those within each tier who achieved at least 30% improvement on each CAT-MH index (ie, CAT-DI, CAT-ANX, and CAT-SS) were calculated among participants considered at least moderately engaged (ie, completed at least four sessions/lessons).

The project was approved by the UCLA Institutional Review Board (approval numbers 17-001938, 16-001395, and 17-001365).

Results on the subsample (n=79) who completed the satisfaction survey are reported in Table 3. In sum, participants across tiers 1-3 rated their care as highly logical (scores ranging from 6.65 to 7.60 out of 9), that the program succeeded in meeting their expectations (scores ranging from 3.47 to 3.95 out of 5), that they were satisfied with the program (scores ranging from 5.24 to 5.90 out of 7), and that they would recommend it to a friend (scores ranging from 5.47 to 6.14 out of 7). Tier 0 participants also rated these categories highly but scores should be interpreted with caution, given the small sample size in tier 0 (Table 3).

This preliminary study demonstrated that STAND, a technology-assisted and scalable system of care, is feasible and effective on a college campus. STAND facilitated the deployment of a significant number of suicide risk assessments as needed in real-time, a critical index of risk prevention. As part of the STAND model, a large number of coaches (n=530) and clinicians (n=47; in psychiatry and psychology) were trained to provide support for or deliver evidence-based treatment for anxiety and depression, respectively. Care was delivered to 516 UCLA students (ie, those who completed at least one session of therapy/digital therapy lesson), who, on average, experienced significant reductions in their anxiety and depression symptoms across all tiers. Participant acceptability on the satisfaction survey revealed moderately high ratings of satisfaction with participants’ assigned tier. Importantly, the stratified model of initial triaging to the appropriate level of care was successful in that the vast majority of participants showed significant improvement in their originally assigned tier. Indeed, only 5 participants required a higher level of care during their initial participation in a lower tier.

Although only a small percentage of participants who completed the initial CAT-MH screening assessment and were offered care actually initiated care in one of the tiers (516/3580, 14.41%), there is a wide range of uptake reported in similar digital interventions, with the rate in this study falling within that range [5,49,50], and promising outcomes emerged among those who did participate. Additionally, a prior analysis of racial/ethnic differences in STAND treatment uptake midway through the recruitment period found that students from ethnic/racial minority groups were not less likely to enroll or engage in treatment compared with non-Hispanic White students (see [51]), indicating that STAND has the potential to reach and be acceptable to a diverse student population, which is important for our current work that aims to implement STAND in underserved communities. Of note, although significant improvement over time was observed across all outcome measures within each tier, a relatively lower percentage of participants in tier 2 showed clinically meaningful change across outcomes than tiers 1 and 3, suggesting that there may be some participants who were assigned to tier 2 who may have benefited more from a higher level of care. Our current work is examining whether including other variables (eg, life stress, trauma, demographics) in multivariate predictive models alongside symptom severity can improve upon and personalize tier triaging to improve clinical outcomes.

Taken together, these findings support the feasibility, acceptability, and preliminary effectiveness of STAND; and this initial demonstration of the STAND model of care provided an opportunity to identify areas for improvement and refinement of the model, including increasing engagement and retention in care.

The data presented represent the initial implementation of this system in an open trial. Therefore, conclusions about the effectiveness of STAND compared with another treatment cannot be drawn. Further, as is common practice with initial implementations that aim to iteratively refine and improve upon methodology, some minor data-driven changes were made throughout the course of the study (eg, changing CAT-DI categories based on incoming psychometric information, shifting from group to individual coaching sessions). Another limitation was the low response rate, particularly in tier 0 and following the ThisWayUp intervention in tiers 1 and 2. This low response rate precludes our ability to evaluate the longer-term outcome in the lower tiers. Finally, with regard to booster treatment during the monitoring period following the initial acute phase of care, in the present demonstration of STAND, we opted to provide the digital therapy to all participants who had completed tier 3 treatment. However, we did not track usage during this preventative/maintenance phase. Future work we have planned will more precisely identify those who may need this additional booster and at what point in their trajectory. Finally, we did not systematically collect demographic data (eg, gender, race/ethnicity) on coaches and clinicians, which would have been interesting to examine both descriptively and correlations between participant/clinician match on demographic factors and clinical outcomes/engagement. Our current work is now collecting these data on coaches in tier 2 as well as conducting a randomized trial to examine the potential impact of racial/ethnic matching between coaches and participants on engagement and clinical outcomes.

Two notable observations in this initial implementation of STAND were the low initial uptake of STAND services, and the relatively modest ongoing engagement in the digital therapy tiers. With regard to the former, the vast minority of students who were screened and eligible to receive services in STAND actually engaged in care. Despite our multipronged approach to recruitment, and the possibility that a lack of financial incentive to participants in the study may have contributed to some extent in low uptake, improvements to our recruitment strategies are clearly needed. Specifically, additional research is needed to identify recruitment strategies (possibly tailored to each new population that adopts STAND) that provide messaging that resonates with the student population. We are currently partnering with social marketing experts to use inclusive, nonstigmatizing messaging and improve upon our outreach efforts; and current research projects underway are seeking feedback from students who did not engage in care to identify barriers and develop solutions.

With regard to the latter, consistent with prior evidence [49,52], retention in digital therapy was somewhat problematic, and utilization of coaching support was particularly low. This is a ubiquitous challenge in digital therapy research for which creative solutions are being developed and evaluated (eg, the use of gamification, avatars, personalization, feedback, and individual support; see [53] for a review). Although our digital therapy approach already incorporates several of these features, we have made several changes to improve retention in digital therapy and coaching, which are currently being implemented and evaluated in ongoing trials. These include automatic scheduling of all digital therapy participants for coaching sessions, rather than making it an add-on option, and personalizing digital therapy by switching from a unified, one-size-fits-all approach to a modularized suite of digital components that are selected using embedded measurement systems to match an individual’s presenting concerns. The change to a personalized approach to module selection is expected to increase retention given evidence for students to prefer tailored online therapy [54]. Current studies are also underway to develop and evaluate engagement strategies (with an emphasis on culturally responsive messaging) and to implement text messaging as a method to increase engagement.

Other changes to our current iteration of STAND include collapsing tiers 1 and 2 into 1 tier (tier 2), with streamlined coach training and supervision, as well as provision of a self-guided wellness digital program in the original tier 0 (now called tier 1). Given that we found such low engagement in tier 0 (the monitoring-only program), we were unable to analyze data collected from tier 0 participants. Possibly, students did not see the benefit of participating in a monitoring-only program when they had minimal to no symptoms (and were not being financially compensated). Our expectation is that by providing digital prevention tools framed as stress management and wellness, students with no symptoms will be more interested in participating in STAND.

The need for effective and scalable treatments for depression and anxiety is enormous given the high (and rising) rates of these disorders and pervasive lack of availability of accessible, evidence-based care. The Screening and Treatment of Anxiety and Depression (STAND) system of care was developed to address this need. STAND is based on the principles of stratified stepped care, adaptation of care in response to continuous symptom monitoring, and routine detection and prevention of suicidality. The STAND system of care was tested in a large sample of UCLA students, where it led to significant reductions in depression and anxiety symptoms and suicide risk. Efforts are underway to continue to refine and enhance the STAND system of care, while increasing its reach, with an emphasis on reaching underserved, diverse populations, conducting randomized clinical trials to identify the optimal ways to deliver STAND, and developing a pathway for sustainability of STAND as it is rolled out on a larger scale. These sustainability efforts will include stakeholder input and economic analyses to identify the specific resources needed to carry out each component of STAND, including ensuring that a sufficient clinical and coaching workforce is available and trained to fidelity in partnership with a college campus that wishes to implement STAND.