Eligibility criteria

Study design.

Rapid reviews are a simplified systematic review process that is used to facilitate a more timely knowledge synthesis [27]. A rapid approach is deemed suitable in instances where timely access to health information is required to inform decision making on pressing matters [27,28]. Given that SCMs have already been implemented in numerous PSIs across North America [8,11,12], use of a rapid review process to synthesize existing knowledge will provide timely guidance. To facilitate the rapid approach, the sole focus was peer-reviewed literature. Specifically, research articles and protocols reporting on multicomponent, mental health or substance use interventions for PSS based on a SCM, stepped approach, or organization of delivery of resources/services based on individual mental health needs were included. Definitions of these terms are provided in Table 1, and were developed using the literature in consultation with team members, including two mental health researchers, a clinician-scientist and an undergraduate student. To incorporate a student-perspective, the student also played a large role in the data extraction and interpretation.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Definitions Used to Guide Scoping Review.

https://doi.org/10.1371/journal.pone.0319473.t001

Interventions featuring individual counselling alone, delivery of intervention components based on preference alone or individual characteristics unrelated to mental health needs (e.g., character traits [30]) were excluded, as were interventions for which it was unclear if the tailoring was based on individual mental health needs. One intervention was excluded because it was not clear whether the escalation process for addressing higher risk was part of the intervention design or only included for mitigation as part of the trial [31]. To facilitate a more rapid review, Personalized Normative Feedback interventions were excluded when that was the only method of delivering intervention components as this literature has already been extensively reviewed [32–34].

Information sources

Study selection process

The results of the search were uploaded to Covidence, an online program that aids in data extraction and management, and collaboration among team members [35]. Using a pair of independent reviewers (SB & CA), record titles and abstracts were screened for inclusion; conflicts were resolved through consensus. Full-text reports were reviewed to confirm inclusion based on information from the title and abstract, and to screen records with an available abstract. References were found within the full-text reports. When multiple papers were available on the same intervention, the most definitive study reporting on intervention effectiveness was used for extraction, although any background information on the intervention itself was assessed if pertinent. If a definitive study was not available, the pilot study or protocol was used.

Data collection and management

Forms to facilitate data extraction within Covidence were developed by the team and pilot-tested before real data was captured. Using a pair of independent reviewers (CA & RH), data was extracted and verified by a third reviewer (SB). Data extracted included study characteristics (e.g., date, country), population information (e.g., inclusion criteria), intervention details (e.g., mental health issue addressed, modality, mechanism for matching to mental health needs), evaluation framework used (e.g., study design, outcome measures, sample size), and student involvement.

Data synthesis

The study characteristics were summarized using tables and described narratively.

Intervention details

SCMs: The 7 SCMs identified ranged from two to nine steps [8,12,50–54]. Five interventions had a specific algorithm for stepping up based on symptom measures [50–54], one intervention stated that the decisions about steps are made as a collaborative process between the counsellor and student [8], and one intervention indicated that treatment planning occurred during an initial consultation [12].

Adapted Stepped Care and Stepped Care 2.0 are integrated SCMs involving an entire system of campus-based mental health services addressing a range of issues [8,12]. The interventions included eight [12] and nine [8] steps ranging from low-intensity interventions such as online information to outpatient therapy in Adapted Stepped Care, and psychiatric consultation and case management in Stepped Care 2.0. Neither intervention appeared to have an objective algorithm for stepping up but rather decisions were made on an individual basis with the student in consultation with a counsellor.

Screening and Treatment for Anxiety and Depression (STAND) is a stratified SCM with four tiers of care that can be adapted based on continual monitoring of symptoms [54]. Tier 0 includes monitoring for those in the normal range of symptoms, tiers 1 and 2 offer online therapy with certified support for those with mild to moderate symptoms, and tier 3 comprises clinical care from psychologists and psychiatrists for those with severe symptoms.

COMPAS-S is an algorithmic intervention integrated into a local administration of the World Mental Health Surveys International College Student Initiative (WMH-ICS) to identity students at risk for suicidal behaviour [53]. Those reporting 12-month suicidal ideation are provided access to a safety planning app and those identified as high risk by a predictive algorithm or by reporting 12-month suicide attempt are contacted by telephone within 24 hours of completing the survey.

Two interventions are based on the ES[S]PRIT SCM model for sub-clinical eating disorders [50,52]. This model was implemented as part of the suite of campus-based mental health services and contained four steps: screening, online program participation, virtual therapist contact, and in-person counselling referral [50,52]. Decisions to step up are algorithmic, based on objective clinical indicators collected through weekly symptom tracking (e.g., body mass index, bingeing).

The last SCM intervention was designed for PSS mandated to an intervention due to violating campus alcohol policies. This intervention contained two steps: manualized brief advice with a peer counsellor for all participants, and a manualized Brief Motivational Interview provided by a professional counsellor with a personalized report for high risk drinkers [51]. After 6 weeks, PSS assessed to be at a high risk were assigned to the higher step based on an objective decision rule informed by data on drinking episodes and alcohol-related problems.

Stepped Care “Lite:” The 13 interventions in this category are characterized by their ability to flag PSS who require higher intensity interventions and generally contained two “steps.” In 11 out of 13 interventions, this ability to “flag” higher intensity needs involved some type of ongoing symptom tracking. For example, in three interventions, therapist contact was initiated or stepped up [17,18,55] when symptoms were detected to decline. In eight interventions, a personalized list of campus-based resources [21,41,56] or referral to university counselling services was provided [16,36,57–59] based on symptom profile or those scoring in high-risk ranges. In a more unique style, one intervention focusing on alcohol use included two “stages.” Stage 1 contained personalized normative feedback offered to all students. Stage 2 provided those identified as high risk based on bi-weekly self-monitoring surveys with either a resource email, or an invitation to chat with a health coach [19]. Finally, one intervention described as an “alternative” to a SCM, altered the modality of a Brief Motivational Interview for first time violators of campus alcohol policies requiring higher-intensity care. In one step, a Brief Motivational Interview was provided in a group format and in step two, an individual session was offered for those identified as high risk [20].

Matching to Mental Health Needs. The 30 interventions in this group were characterized by containing a mechanism that could match intervention components offered with the mental health needs of the individual PSS. In 19 interventions, the matching process was facilitated using some type of symptom tracking. The most common intervention design in this category (n = 14) involved providing tailored feedback and information based on assessed symptom profiles [37–39,42,43,46,47,49,60–65]. This was achieved most often within the content of the website itself, but in the StudentBodies interventions, the virtual feedback was provided by a human group moderator [37–39] and in iCare Prevent, Best-U and Space from Depression and Anxiety interventions, by a virtual human “e-coach.” [46–48,66] Also common within this group were interventions that provided activities, activity suggestions or specific web content that matched assessed symptom profiles (n = 12) [40,44–49,61,67–70]. Again, usually this was done within the content of the website, but in three interventions it was achieved via “e-coaches” or “supporters.” [40,69,70] To improve therapy progression, an app called elona therapy was used alongside individual cognitive behavioural sessions; the content of the app was tailored by the therapist according to the student’s diagnosis and individual mental health needs [71]. A sub-group of interventions within this category used text messages to achieve matching based on mental health needs (n = 5); three used text messages to address indicated or predicted urges to smoke [72–74] and two used text messages to address actual drinking behaviour and consequences [75,76].

Symptom tracking

In total, 34 interventions incorporated some type of symptom tracking. However, how it was incorporated varied greatly with respect to timing of assessment. A total of nine interventions used daily or more frequent tracking [18,36,44,45,62,63,66,72,74], of which four explicitly mentioned Ecological Momentary Assessment (EMA) [44,45,62,72]. Another eight interventions used a weekly or biweekly schedule [16,17,19,38,50,52,54,55], while three used tracking at each session [57,58,77]. Finally, 13 interventions offered tracking that appeared to be optional and/or could be used at any time desired by the participant. All but two interventions [17,55] tracked using technology, using either in-app or web-based assessments.

Evaluation methods

See Table 2 for an overview of the evaluation methods used for the SCM and stepped care “lite” studies, including the study population, study design, sample size and outcome measures used. Similar information can be found for the third category of studies in S1 Table. There was great heterogeneity in study design, sample size, and outcomes used. Evaluation methods for each study type are discussed below.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Summary of Stepped Care Model and Stepped Care “Lite” Studies.

https://doi.org/10.1371/journal.pone.0319473.t002

SCMs: Only one SCM intervention was evaluated using a design that involved randomization. Specifically, Borsari et al. used a repeated measures design to assess alcohol use outcomes at 3, 6 and 9 months post intervention among those who received step one only (low risk), and those assigned to step two (high risk), who were randomly allocated to receive a BMI or assessment only [51]. By comparing three groups, those at low risk, those at high risk who got the appropriate high-risk intervention, and those at high risk who did not get the appropriate high-risk intervention, not only is the dose-specific response evaluated, but what happens when the hypothesized correct dose is not given is also assessed.

The Adapted Stepped Care [12] and Stepped Care 2.0 [8] models were both evaluated using a system-level pre-post design with focus on service utilization (e.g., number of clients, appointments) [8,12], client satisfaction [8], and step-specific outcomes (e.g., change in scores across psychotherapy sessions) before and after the model was implemented [8,12]. The ES[S]PRIT model was assessed using case study designs focusing on implementation based on utilization data, and average symptom levels for those assigned to each step [50,52]. In the Irish-adapted version [52], utilization patterns of moderate and high-risk participants were compared according to how the program was actually used (e.g., whether moderate-risk participants primarily used the lower intensity interventions, and high-risk participants mostly used the higher intensity interventions). Follow-up outcome data was not included in the evaluation design of either version of the model. Each tier from the STAND intervention was evaluated separately using symptom change scores [54]. Finally, in COMPAS-S an intervention and control cohort from the WMH-ICS were compared to determine if those identified as at risk at baseline in the intervention cohort were less likely to remain at risk at the 12-month follow up than the control cohort [53].

Three SCM studies mentioned validating and optimizing decision rules [50,53,78]. The ES[S]PRIT model decision rules for step assignment were validated using symptom data of those invited to participate in higher intensity interventions, and those not [50]. Borsari et al. conducted a follow-up study to assess which individuals responded best to each step, with the goal of understanding how the intervention could be further developed by varying the content and intensity to better meet the needs of individual students [78]. The COMPAS-S algorithm was validated using a different cohort from the WMH-ICS [53].

Stepped Care “Lite”: Five studies within this category applied randomization [16,17,19,36,59]. Three studies focused on platform usage/engagement as a proxy for intervention dose [16,18,41]. For example, one study determined whether use of modules within the app suggested based on daily mental health assessments was associated with increased engagement and improvement in screening outcomes [18]. Sub-analysis by risk levels according to levels of engagement with the app (e.g., assessments done, links clicked) was also undertaken in the feasibility study of MindYourHead [41]; however, outcomes by level of engagement were not assessed. In the protocol for the Health Mind Network, it was stated that change in outcomes over time will be assessed by app usage (e.g., number of sessions completed). One study used a system-level pre-post design to assess the number of students using campus-based services based on platform-based recommendations before and after the app IntelliCare for College Students was launched, but did not use individual-level data to assess utilization by risk level [56].

One study in this category mentioned validating and optimizing decision rules. Patrick et al. conducted a follow-up study on their staged intervention for alcohol using a Q-learning algorithm to estimate the most optimal decision rules for tailoring variables [79].

Matching to Mental Health Needs. Two studies in this category focused on level of engagement with the platform as a proxy for intervention dose [68,72]. Evaluation of an app called Nod that provided tailored activities based on mood assessment looked at outcomes by level of engagement [68]. Evaluation of the BASICS-Mobile app looked at outcomes by the number of modules received [72]. In another study that provided tailored health information based on drinking profiles, drinker type--which is correlated with the profiles--was compared by study outcomes [60]. In the evaluation of Space from Depression and Anxiety, an unguided version of the app that contained no tailoring was compared with a guided version to isolate the effect of tailoring from the core content of the app; however, outcomes associated with individual dose within the guided condition were not assessed [48]. Decomposition was also used by Levin et al., who decomposed the effect of EMA alone compared with EMA with the other tailored intervention components to isolate the effect of symptom tracking [44].

Student-relevant outcomes

Table 2 provides the mental health outcomes used across the interventions. We were interested in how many studies incorporated outcome measures related to being a student. In total, nine interventions used this type of outcome measure, most often in relation to academics, with specific examples including: academic impairment [37], academic self-efficacy [43], perception of academic functioning [17], GPA [16], enrolment status [16], and intention to stay enrolled [68]. Other examples of student-relevant outcome measures include campus belonging [48,68], fraternity/sorority membership [62], and intention to pledge for “Greek Life.”[19].

Student co-design

Student co-design was most often reported on in the third category of interventions--matching to mental health needs, where nine studies applied it to some extent [43,46–49,62,63,65,68]. Three of the stepped care “lite” interventions mentioned involvement of students in some aspect of co-design [41,56,59]. None of the SCM interventions reported on student co-design. In all cases, it appeared that students were involved in initial design or adaptation work. One intervention indicated that a student “leadership group” was involved [43]. Focus groups were the approach most often used to involve students [47,48,63,65,68].

This rapid scoping review was undertaken to better understand the landscape of SCMs and other related approaches designed for PSS, reflecting the increased interest in SCMs for addressing what has been framed by some as a “mental health crisis on campus.” [4,80,81] Consistent with the review by Berger et al. of SCMs for youth up to 2020, few interventions based on an SCM were found for PSS as a unique population [9]. Specifically, seven interventions were based on a SCM as defined by the intervention designers themselves. However, even within this small group, heterogeneity was observed. Two of the SCMs were integrated system-wide models including an array of mental health services spanning 8 or 9 steps [8,12]. The other five were narrower in scope, made up of a discrete set of interventions across 2 to 4 steps, each addressing particular mental health issues, including depression and anxiety [54], an eating disorder [50,52], problematic drinking [46–49,51,56,62,63,65,68], and suicidal behaviour [53]. The literature is therefore both limited and lacking a consistency in how SCMs are envisioned for PSS.

By applying a wide scope to this review, we were able to identify 13 interventions that, while not identified as SCMs by the intervention designers themselves, could be characterized as stepped care “lite.” Almost all interventions identified in this category used a website or app as the first “step”, with higher-intensity steps involving direct access or referrals to existing services offered on campus. While spanning fewer steps, these interventions address the requirement of increased access and efficiency, as well as the ability to meet the diverse needs of students. Further, by incorporating symptom tracking, they provide ongoing assessment of mental health needs and the ability to flag those who require higher-intensity care in real time. However, while there is great potential for using technology-driven approaches such as apps, there remains questions as to how students engage with them over the longer term, with studies showing low adherence to app use over time [82,83]. Interviews with PSS have revealed ambivalence toward apps, on the one hand believing in the potential for technology to help but on the other, PSS question the extent to which individuals can effectively manage their mental health on their own [84]. Concerns about privacy and credibility have also been raised [83]. Supported interventions that include therapist check-ins may be more effective, suggesting that integrating technology-driven approaches with traditional counselling services may be the best approach [85]. Therefore, while stepped care “lite” interventions that incorporate technology likely will not replace broader, system-wide models, they may fit into the general landscape of stepped approaches that serve student populations, such as those on waiting lists for services [45], those who could benefit from briefer sessions with a counsellor accompanied by an online platform [55], or those who are not yet accessing mental health services but may require higher-intensity interventions [16].

In a third category of studies, 30 interventions were identified that focused on matching intervention components to the mental health needs of the recipient, but without then changing the intensity of care within the intervention itself. These interventions fit within the landscape of SCMs and stepped approaches, as a key element of SCMs as described by Cornish et al. is the ability to organize delivery of services/resources according to individual mental health needs along with readiness, functioning and capacity to engage [10]. We were interested in how processes for matching to needs might work, and whether evaluation models had been designed to capture outcomes associated with having received an intervention tailored specifically to that individual. Most commonly, these interventions involved providing tailored feedback and information based on symptom profiles developed using data from symptom tracking (e.g., [43]). Several interventions in this group also offered “e-coaching” to increase engagement. Low intensity interventions that provide tailored content fit into the stepped care landscape by offering information and advice that fits the unique needs of the individual. However, striking the right balance between presenting helpful and engaging material and exacerbating the problem with unnecessary embellishments must be considered. “Digital burnout” has been identified as a concern for apps offering notifications [86]. In a similar vein, research based on the MindLAMP app demonstrated that greater engagement was fostered by having PSS interact with their own data (e.g., symptom tracking) rather than by sending encouraging messages [87]. While it may feel like with ever-newer technology anything is possible, students themselves have indicated that simplicity is important [83], which means avoiding the add-on features and just focusing on the tailored content.

When considering the literature reviewed, we have three recommendations for moving forward research on and implementation of SCMs for PSS. Similar to the call made by Salmon and colleagues for a consistent definition of SCMs for youth [15], we are recommending that a definition of SCMs for PSS needs be developed that is unique to the post-secondary context. This definition must speak to both the unique circumstances of the post-secondary environment, and the needs of students as differentiated from youth more broadly. As for the latter, while PSS and youth are overlapping populations, students experience unique stressors that include those related to achievement, campus culture and financial pressure [88]. Also distinctive from youth populations is the substantial proportion – upward of 30% in some regions -- of PSS that are international students [89]. Providing care to those studying in a country for a limited period of time is complex and requires culturally-adapted approaches [90]. As far as the post secondary environment goes, the definition needs to reflect the fact that campuses are not part of a formal health care system, and generally lack the ability to provide comprehensive assessments, ongoing care, and specialized services such as psychiatry [11]. Therefore, a definition of stepped care developed for youth accessing formal mental health care systems will not fit the circumstances of students accessing care on campus. In addition, with the movement toward whole university approaches that embed considerations of mental health into all aspects of post-secondary life [91], and an increasing focus on salutogenic perspectives on student mental health [92], the definition must be broad enough to consider prevention and population health more broadly. From a student perspective, access to and waiting for care are big concerns [93]. We must therefore consider the margins of care as much as the system of care itself when developing a consensus definition of SCMs for PSS. Future research should consider using a Delphi consensus study, including students, to determine the core components of a definition of SCM for PSS and their evaluation outcomes. Recently, members of our team undertook a Delphi consensus study to understand priorities in student mental health research from a global perspective [94]. Key to the design of the study was including separate groups of professionals, academics and students so that the voice of any one group did not dominate. Results of this current review could be used to develop core components of a definition that could be rated as part of the consensus-building process. While not all interventions identified in this review may be included within a consensus definition, it will be important to consider each with respect to why they do or do not fit.

A second recommendation is that PSIs need not only to undertake evaluations of their locally implemented SCMs, but also draw on their internal research expertise to develop robust designs that provide high quality, publishable data. Without accrual of evidence, it will not be possible to assess the efficacy and provide support for the increasing implementation of this model in PSIs. Following existing recommendations, SCMs should be evaluated as a whole, not at each step separately [15,24]. Another way to think of this is in terms of dose-specific effects, the dose being the specific intervention components (or steps) exposed to. While some attempts were made to assess the overall model using aggregate pre-post data [8,12], or engagement with the website or app as a proxy for intervention dose [16,18,41], the evaluation models reviewed were generally not geared to look at individual-level outcomes capturing the overall effect of the model. This is not an easy undertaking, and will likely require highly pragmatic designs that are co-created with PSS and service providers to integrate practical considerations of the actual setting [95]. Longitudinal cohort studies that follow students first entering a PSI across multiple years, such as the SWANS cohort study [96], could be used to capture use of an initial intervention, and any subsequent intervention use within an SCM across time. Ideally, PSS would be followed past graduation, to understand longer-term outcomes and the transition from campus-based into adult mental health care systems. Additionally, while randomization is the gold standard for the evaluation of interventions, it is not always clear how to incorporate it in the assessment of stepped care. The two-step SCM described by Borsari et al. was evaluated using a design that involved randomization at step two only. Since all participants received step one, this isolated the effectiveness of stepping up to among those who were determined high risk [51]. Patrick et al. used a sequential multiple assignment randomized trial (“SMART”) design involving two randomizations: timing of stage one and, for high-risk participants receiving stage two, the type of intervention provided [19]. One idea for incorporating randomization-like conditions for models with many steps is to make use of a natural experiment. This general approach was undertaken by Hasking et al., who compared two cohorts of students completing the WMH-ICS – one that received an SCM intervention and one that did not [53]. The using of matching techniques could enhance this design to make the cohorts even more comparable. For this work to be feasible, PSIs need to prioritize high-quality evaluation of their own mental health services. In doing so, they should consider how strategic partnerships between academics specializing in disciplines related to student mental health and the professionals involved in the delivery of student services can be developed. At our own institution, a mental health researcher has been embedded into the administrative student mental health team to support high-quality evaluation projects.

The third recommendation is that students must be purposefully and authentically engaged in this work. In our review of the literature, we found that student co-design was often not present, or at least not often reported. When students were involved, the engagement was generally confined to the initial design/adaptation or pre-testing phases rather than being iterative, and little detail was provided to assess the quality of involvement. As student co-design becomes more widely implemented in student mental health research, it will be interesting to assess whether studies that more thoroughly and effectively engage students have better outcomes than those that have minimal student involvement. From a broader perspective, in addition to being involved in studies of SCM interventions, students should be key to helping to define the core components of SCMs for PSS. From a student perspective, this is essential to ensure that when PSIs talk about implementing an SCM, that students know what they are referring to based on a verifiable set of core components. This will also help guide PSIs to implement SCMs that contain the necessary components as endorsed by students themselves. Students also need to be included in the discussion of what outcomes should be focused on for evaluation; this approach had been used by the YouthCan RCT of a stepped care pathway for youth [95]. In our review, we observed that symptom measures were used often, as well as measures like satisfaction and utilization. However, outcomes related to being a student specifically were often missing. We were surprised to find that only a small minority of studies had incorporated functional outcomes such as academic retention and success. While we do not know if students prefer these types of outcomes to the more symptom-focused ones, it is important to find out. Models for youth engagement, such as that used by YouthCan including both high and broad engagement strategies [97] should be implemented in post-secondary settings. A recent review of models for engaging PSS in campus-based mental health research provides specific guidance how this work can be done within a post-secondary context specifically [98]. For example, from a relational perspective, creating a comfortable and safe environment is key, as well as mitigating power imbalances between researchers and the students [98]. From a process perspective, it is recommended that diverse and multiple methods for collecting feedback are used, small group discussion is facilitated, and clear roles and responsibilities for the students are outlined [98].

This study is not without some limitations. First, as part of our decision to support a rapid approach, we did not review the grey literature. We therefore did not include results from dissertations or research reports not published in the peer-reviewed literature, somewhat limiting the comprehensiveness of the review. Second, since we wanted this review to be broad, we included search terms that would capture features of SCMs over and above the term stepped care specifically. This enabled us to find interventions that were not described as SCMs but contained features consistent with this model. However, it is possible that we did not include enough related terms to rule out the possibly that we missed some interventions that would have fallen within the stepped care “lite” category. Third, the review was limited to English language only, and with evidence of interventions being implemented globally, we likely have missed examples of work being undertaken in some regions. With students moving from region to region to obtain post-secondary qualifications, even PSIs within English-speaking regions need to draw on global data to develop approaches that support students from a variety of cultural backgrounds. Most of the studies reviewed here came from the US, and only one study focused on international students specifically [41], which means that our understanding of SCMs for PSS comes from a limited perspective. It is possible that key features of SCMs may not be equally received by all cultures. For example, the focus on collaboration between care provider and client in SCMs may confront cultural preference for paternalistic care [99]. Future research, including that recommended to develop a consensus definition of SCMs for PSS, should be designed to include a global perspective to ensure this work is applicable to all PSS. Finally, while we put much emphasis on the need to include students in this work, our own inclusion of students was not as extensive as possible. We acknowledge this and are developing a plan to work with students as we move this project forward.