University students’ mental health problems are highly prevalent globally [1]. In Finland, 36% of university students reported high levels of psychological distress and 24% reported increased loneliness in the 2021 Finnish Student Health and Well-Being Survey [2]. Anxiety and depressive symptoms at the start of university studies are often persistent and related to poorer academic performance and lower engagement [3]. The need for accessible, cost-effective mental health services in universities is substantial [4].

Loneliness (the discrepancy between one’s actual and desired social relationships) has been recognized as one of the key risk factors for numerous adverse psychological and somatic health outcomes among students and youth [5] and in various population-based samples [6,7].

Having positive relationships and feeling socially connected is also essential for academic performance. Loneliness, social isolation, and low social status impact intellectual achievement [8], whereas engaging in the academic environment reinforces individuals’ learning activities [9].

Mental health problems are more prevalent among many minority groups, who experience psychological symptoms more often than the rest of the population [10-12]. Several studies have also found that belonging to more than one minority group increases the risk of stressful life situations, mental health problems, and thus the need to use mental health services [13,14]. Belonging to minorities may also increase the risk of social isolation. According to the School Health Promotion study [12], young people belonging to gender and sexual minorities show elevated levels of loneliness, exclusion, and difficulties in receiving help in Finland. In addition, young people with immigrant backgrounds and young people with disabilities face considerable risks of loneliness [15]. These difficulties are largely caused by minority stress, which in turn has been linked with experiences of discrimination and unequal treatment [12,16].

With this background, interventions targeting loneliness and supporting social connectedness, inclusion, and equality could be a significant way to approach the student mental health crisis. The Groups 4 Health (G4H) intervention studied in this protocol aims at improving well-being by supporting individuals’ social group memberships that are compatible with their social identity [17].

G4H builds on the social identity approach to health (SIAH) [18,19] and the social identity model of identity change (SIMIC) [20,21]. The SIAH [18,19] comprises 2 related theories, the social identity theory [22] and self-categorization theory [23], claiming that individuals’ internalized group memberships have a significant role in their sense of self and that shared identities are important for productive social interaction. Therefore, belonging to multiple social groups that align with the individual’s identity is crucial for a sense of purpose and belonging, as well as for receiving support from others [18]. The role of group memberships is further emphasized during transitional life events such as moving to another country or entering a university, which may be stressful and may also affect an individual’s access to their social groups. The SIMIC further states that during stressful and transitional life situations, well-being can be enhanced by both managing and strengthening the individual’s social group memberships before the life change (the social identity continuity pathway) and by becoming a member of new social groups that align with the individual’s identity (the social identity gain pathway) [20,21].

The fact that G4H addresses major life changes and social identities makes it particularly relevant in the university context. Many students experience major changes, such as relocating to a new country or city, integrating into new communities, and leaving familiar ones behind. For minority students, finding new social groups that align with their identity can be even more challenging, emphasizing their need for support.

This protocol provides a unique opportunity to study the G4H in the university context with a multimethod approach that integrates both quantitative and qualitative methods. By implementing this approach, we aim to capture not only the effectiveness of the G4H intervention and the participant- or facilitator-related factors that moderate the effectiveness, but also the perspectives of minority students participating in the intervention, which is an aspect that has been overlooked in previous research. The inclusion of qualitative methods is crucial, as it allows us to gain deeper insights into the experiences and potential challenges faced by these students, enriching our understanding of the intervention’s impact.

Additionally, neither the cost-effectiveness of the intervention nor its effectiveness on academic performance has been explored in previous studies. Both of these analyses are essential for assessing its suitability and adaptability in promoting student well-being. Furthermore, to our knowledge, the effectiveness of the G4H has been assessed only in Australia in studies organized by the developers of the intervention. These represent clear gaps in the existing literature that this protocol aims to address.

The aim of this study is to investigate the effectiveness of the G4H intervention for university students’ well-being using both quantitative and qualitative methods. Our study includes 2 active comparators: intervention groups facilitated by the University of Helsinki (UH) study psychologists and an online intervention course focused on well-being and study skills. A control group of students who are not participating in any active well-being services at UH is also included.

Our first objective is to investigate the effectiveness and cost-effectiveness of the G4H intervention in (1) supporting well-being (most importantly by reducing loneliness and depression) and (2) supporting academic performance among university students. The second objective is to explore which factors moderate or mediate the effectiveness of the interventions. The third objective is to focus on the G4H intervention’s inclusiveness and the capacity of the G4H intervention to promote equality via qualitative methods.

Our research questions (RQs) are as follows:

Compared with the no-intervention control group, we hypothesize that all active interventions reduce students’ loneliness and depression (RQ1) and improve students’ academic performance (RQ2). However, participants in the G4H intervention will experience a more long-term reduction in loneliness than those in the other active intervention groups [24]. We hypothesize that treatment alliance [25,26] and higher fidelity to the G4H manual mediate the interventions’ outcomes (RQ4). Regarding facilitators’ characteristics, we hypothesize that their greater self-rated skills and enjoyment in therapeutic work as well as agreeableness, openness, and engagement in personality traits and interpersonal relating will be positively associated with treatment alliances and intervention outcomes; and, conversely, greater difficulties and negative feelings in therapy work will be negatively associated with treatment alliances and outcomes [27-29]. Regarding motivation (RQ4), we set our hypotheses based on the self-determination theory [30] and the integral model of treatment motivation [31]. We hypothesize that autonomous motivation to participate in the intervention predicts greater engagement and increased effectiveness, whereas controlled motivation predicts lower engagement and effectiveness. We further hypothesize that engagement partially mediates the relationship between motivation and intervention effectiveness. No preliminary hypotheses are posed regarding RQs 3 and 5.

We conduct a parallel arm nonrandomized controlled trial with a 1:1 allocation ratio. We compare 3 intervention arms to a no-intervention control group as follows:

The study design and timeline are presented in Figure 1. The study follows the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines.

To collect the quantitative data, we use paper-and-pencil and electronic versions of the questionnaires. Those in the active intervention groups fill out the questionnaires before or after the intervention sessions if the questionnaires are provided electronically and during the session if the questionnaires are provided on paper. Those in the no-intervention group fill out the questionnaires based on the email reminders of the electronic data collection system (time interval between reminders is the same as the interval between sessions in the active intervention groups).

Electronic data collection is implemented in the Research Electronic Data Capture (REDCap) system (Vanderbilt University) [41], provided by the UH for collecting sensitive research data electronically. In the case of the TBWS intervention, data collection from time points before and immediately after the intervention is partly integrated into the online course platform, and the students can respond either via the course platform or via the REDCap system, while data collection from week 3 and the 1- and 3-month follow-ups is implemented only in the REDCap.

Before the analysis phase, the data validity is confirmed via range checks, missing value checks, and outlier detection. Potentially erroneous data are checked against the paper and electronic forms, and other necessary sources, and corrected if needed. All modifications are documented.

Participant retention is promoted by reminding the participants during the group meetings to fill out the questionnaires. All participants receive up to 5 reminder messages from the electronic data collection system if they do not complete the necessary questionnaires on time. Retention is also promoted by the possibility of receiving written feedback on the questionnaire results after participation. Participating students also receive movie tickets as compensation for completing all follow-up questionnaires. Adherence to the G4H intervention is monitored by documenting the number of sessions in which each student participated.

All the quantitative methods used are presented in Table 1, which also includes the measurement time points and study arms from which the data are collected. Multimedia Appendix 1 includes a table that describes the unabbreviated titles of each scale, scale details, and information on their reliability and validity. Most measures have demonstrated at least acceptable internal consistency (Cronbach α≥0.70) and convergent validity with other similar measures. The few exceptions with lower Cronbach α are either very short scales (4 items) or include a specific subscale that performs poorly.

We will conduct our analysis following the intention-to-treat principle, including all participants regardless of whether they completed the intervention.

The change in the outcome variables over time will be analyzed with linear mixed-model analyses due to the hierarchical data structure. Information criteria are used to assess model fit and determine the appropriate model structure. Mixed effects models can also be fitted for participants with missing data from one or more time points. If needed, other appropriate imputation methods, such as multivariate imputation by chained equations [63], will be used. For this reason, the number of participants included in the analyses is likely to be higher than the number of participants providing data at the last follow-up.

We will explore with logistic regression analyses whether background characteristics and baseline psychiatric symptoms predict study dropout, and necessary means to account for any systematically missing data will be taken. We aim to show that the missing observations of the outcome variable, due to dropout, are conditionally independent from the observed values of the outcome variable given our model for dropping out from the study. If needed, inverse propensity score weighting is carried out so that the treatment effects are conditionally independent from the treatment group assignments and from differing dropout rates given the propensity scores. We consider individuals’ selection into different study groups (arms) under the stable unit treatment value assumption based on their background characteristics, given that we have access to an extensive dataset. The selection of individuals into groups is accounted for by making the estimators double robust, which is a well-known approach to adjust for selection into groups based on background characteristics in model-based estimates [64,65]. This ensures that our results are unbiased and consistent estimators of the treatment effects. Potential effects of the COVID-19 pandemic on the stability and change in students’ well-being will be inspected and considered in the analyses if necessary.

All models will be inspected for normality of residuals, homoscedasticity, and multicollinearity. The outcome variables will be transformed to attain normality if needed. Appropriate interaction terms are added for further subgroup analyses. Potential mediator (eg, therapeutic alliance, motivation, and model fidelity) and moderator (eg, participant and facilitators' background) effects will be inspected in additional mediation and moderation analyses.

In the case of missing questionnaire items, within-person imputation is carried out for responses when less than 50% of the answers are missing per scale (the within-person average of the existing items is imputed to the missing responses). We acknowledge that this procedure may decrease the within-person variation in the data. However, as we discuss later, the prevalence of item nonresponse in the dataset is low.

We will take appropriate measures to balance between Type I and Type II error in the reporting of analyses results. If needed, we will use methods to account for multiple testing such as false discovery rate or new eigenvalue-based methods that provide a more accurate adjustment for nonindependence among tests [66].

We will conduct 3-month follow-up cost-utility and cost-effectiveness analyses. The cost-utility analysis will use EQ-5D-5L–based quality-adjusted life-years as the effectiveness outcome measure [50,67]. The cost-effectiveness analysis will use both primary outcomes of the effectiveness study, namely loneliness measured with the 8-item University of California Los Angeles Loneliness Scale [42] and depression measured with the Patient Health Questionnaire [44] (Table 1 and Multimedia Appendix 1) as the effectiveness outcome measure. The analyses will use a societal perspective by including the intervention costs and costs of students’ health and social service use. The analyses will be conducted with the intention-to-treat principle. The missing outcome measures will be imputed with multiple imputation methods [68]. We will estimate incremental net health benefits between interventions. The estimation of the incremental net health benefits and 95% CIs will be used for statistical methods (eg, ordinary least squares regressions) [69]. We will perform multiple 1-way sensitivity analyses to test the robustness of the economic evaluation results. The results of sensitivity analyses will be presented in a tornado diagram. If interventions show effectiveness, we will conduct decision-analytical modeling of long-term cost-utility and cost-effectiveness beyond 3-month follow-up data.

We explore inclusion and equality, with respect to G4H using three qualitative data collection methods: (1) participant observation, (2) interviews with G4H participants and facilitators, and (3) focus groups with young people who belong to minorities. The aim is to examine challenges and possibilities in the G4H intervention regarding social inclusion and equality and whether G4H promotes long-term social inclusion and equality among the participants. The focus groups target young people belonging to sexual and gender minorities, young people with immigrant backgrounds, and young people with disabilities. These groups represent youth minorities with a considerable risk of loneliness [15]. To date, research on their experiences with different services is lacking [12,70-72]. These groups are also underrepresented in higher education [2,73,74].

The participant observation focuses on the interaction and the practices undertaken in the groups to gain knowledge about how the groups work, what kinds of issues they deal with, and especially how inclusion and equality are accomplished in the groups (are there minority participants, can participants influence the themes dealt with in the groups, or is inclusion and equality taken into account in some other ways in the groups). Following the ethnographic research tradition, the researcher takes part in the activities of the groups instead of observing them from the outside. Depending on their preferences, the study participants (both facilitators and participants in the groups) will be interviewed in a group, in pairs, or individually, in facilities located outside the university (eg, in a study room of a local library).

In the focus groups, young people belonging to minorities are asked to comment on the design or description of G4H in terms of equality and inclusion, with the goal of enhancing inclusivity by listening to their views on the interventions and accumulating knowledge on the fit between the interventions and their specific needs. All the interviews will be tape-recorded and transcribed for analysis.

To ensure the validity of qualitative findings, data triangulation is used via collecting 3 different types of data: focus group interviews, intervention participant interviews, and facilitator interviews. Setting these different data into a dialogue enables a multifaceted understanding of intervention effectiveness, inclusion, and equity. The analyses of qualitative data draw on minority stress theory [75] and intersectional theory [76,77].

The analyses are conducted by following reflexive thematic analysis [78,79]. Reflexive thematic analysis is conducted by following 6 steps that move from familiarization with data, coding, formulating, and refining themes to writing the report in alignment with the key principles of the approach, highlighting reflexivity and in-depth attention to nuance in participants’ meaning-making. If several analysts are involved in the analysis, the goal is to engage in discussion throughout each step of the analysis, to refine the analysis with the help of various analysts’ viewpoints. This approach, therefore, acknowledges that analysts’ interpretations of the data may differ, which is why transparency and a close consideration of the subjectivity of interpretation are highlighted. The first steps of the analysis may be conducted with the help of software (such as ATLAS.ti 24; Lumivero), but the actual analysis requires repeated rounds of reading of the whole dataset by the analysts. The emphasis on researcher reflexivity in this analytical approach further enhances the validity of interpretations and helps ensure that the findings reflect the multiplicity of participants’ viewpoints and accurately portray repetitive patterns in the data.

Figure 2 illustrates the number of participants from the start of the study in March 2022 until the first data freeze in March 2025.

A high proportion of the university students enrolled in the interventions decided to participate in the study. The minimum target sample size (n=150) was attained for the G4H (n=155), TBWS (n=216), and control group (n=203). For the SPGs, the target sample size was not attained (n=125).

The anticipated study completion date for all data collection is by the end of the year 2028. Analyses for RQ1 and RQ3 are anticipated to take place in years 2025 and 2026, analyses for RQ2 in years 2026 to 2027, and analyses for RQ4 and RQ5 in years 2026 to 2028. First results from the study are anticipated to be published in the year 2026.

Figure 3 illustrates the attrition rates by trial arms by the first data freeze in March 2025. The percentages do not accurately indicate the number of study dropouts, as all surveys are sent to participants regardless of nonresponse to previous surveys, allowing them to re-enter the study. The attrition rate has exceeded the anticipated 35% in the SPGs and TBWS. For the TBWS, the bars with black outlines represent the responses to the REDCap survey. The low response rate at the week 3 assessment (during intervention) is due to the survey only being available in REDCap and not integrated into the online course platform as in week 1 and week 7. Based on the attrition rate, we took action to increase retention, including adding incentives for the participants (a movie ticket for completing the surveys).

Table 2 illustrates the participants’ characteristics of the current sample at baseline by trial arm. In all groups, most participants are women (87/125, 70%‐175/203, 86%), not in a relationship (119/214, 56%‐107/154, 69%), and childless (175/208, 84%‐101/113, 89%). The age medians are lower than the age means (median_G4H 25, IQR 22-29 y, median_SPGs 27, IQR 25-31 y, median_TBWS 24, IQR 22-31 y, and median_Control 23 y, IQR 21-27 y).

To determine whether the participants in the trial arms differed from one another in terms of background characteristics, we performed chi-square tests for categorical variables and Kruskal-Wallis tests for continuous variables (based on visual inspection, neither followed a normal distribution) (Table 2). Trial arms differed in age (Kruskal-Wallis χ²₃=40.7, P<.001), gender (χ²₆=5.4, P<.001), relationship status (χ²₆=15.8, P=.02), living situation (χ²₆=12.6, P=.05), and study years in current degree (Kruskal-Wallis χ²₃=125.7, P<.001). These observations highlight the need to address group selection bias.

Item nonresponse (ie, a participant does not respond to an item within a questionnaire) was observed with missing data pattern plots across all groups. First, we examined the outcome variables which are assessed at all time points except week 3: loneliness, depression, mental well-being, general well-being, anxiety, social anxiety, multiple group memberships, and mood. Structural missingness (not all questionnaires were administered to all TBWS participants) was not included. Across all time points (excluding week 3), 94% (2174/2317) of survey responses were complete, 5% (105/2317) included a single missing item, and 2% (38/2317) included multiple missing items. Regarding the shorter survey taken at week 3 that includes the assessment of group memberships, mood, identification with the intervention group, loneliness, and working alliance, 96% (499/518) of responses were complete, 3% (16/518) included a single missing item, and 1% (3/518) included multiple missing items (0‐7 missing items per survey). The level of missingness within given survey responses appears acceptable, especially when considering the length of the longer surveys (55 items, approximately 15‐23 min).

Missingness was more prevalent for the selected background variables (Table 2). Of all survey responses of these selected variables, 79% (581/732) of responses were complete, 19% (141/732) included a single missing item, and 1% (10/732) included multiple missing items. Most often, participants left their number of children unreported. This may be due to a misunderstanding where some childless participants skipped this question instead of reporting zero. However, as the overall missingness in the selected background variables is 3% (172/5856), it is still within acceptable limits (<5%).

Psychological ill-being and loneliness are highly prevalent among university students, often leading to long-lasting negative effects on mental health, social connectedness, and academic performance. To address this growing issue, accessible and preventive services at universities are needed. We investigate the effectiveness of the G4H intervention for university students’ well-being using both quantitative and qualitative methods including 2 active comparators.

We hypothesize that all the active interventions will improve students’ well-being and that the decrease in loneliness will be more pronounced for the participants of the G4H group, in line with earlier studies [17,24,33,37]. We also anticipate valuable qualitative insights into the experiences of minority students. The multimethod approach will provide a comprehensive understanding of the intervention’s effectiveness and the contextual factors influencing its impact.

A key strength of this study is its multifaceted approach, which allows for a comprehensive exploration of existing student well-being interventions at the UH. By involving multiple collaborators, each with unique research interests and expertise, the study can address a wide range of critical questions related to the interventions’ impact. This collaborative framework enables the examination of various mediators and moderators, as well as the integration of both qualitative and quantitative methods. This perspective not only enriches the understanding of students’ well-being but also ensures that these phenomena are not oversimplified. The multimethod design of the study allows participants to express their thoughts and experiences in more depth, enhancing participant engagement and potentially leading to richer data.

Furthermore, the qualitative insights gained from this approach will provide a better understanding of the intervention’s effectiveness—or lack thereof—from the perspective of minority students. This insight is crucial for further developing the G4H intervention in terms of inclusion and equality. Moreover, the study is particularly relevant from an implementation point of view, as it will ensure the suitability and effectiveness of the method in Finland, addressing the significant need for early-stage support within the service system.

This study also has some limitations. The comparator arms (SPGs and TBWS) differ from the G4H slightly in terms of intensity (G4H: 5 sessions, TBWS: 6 sessions, and SPGs: 5‐7 sessions) and delivery methods. This variability complicates direct comparisons between these groups and the G4H intervention, potentially introducing confounding factors that could influence the study’s outcomes. Additionally, the varying data collection methods may introduce mode effects or measurement bias, as well as structural missingness, impacting the reliability and comparability of the data. To mitigate these concerns, we will conduct sensitivity analyses and discuss the implications of these methodological differences in our results.

Furthermore, the lack of exclusion criteria and the nonrandomized design of the protocol may introduce potential biases regarding group composition. However, allowing students to choose which intervention they participate in respects their autonomy and aligns with ethical research practices. This nonrandomized design increases the ecological validity of our findings by capturing the experiences of students in a real-world educational environment, which is important for making informed decisions that can improve services at both individual and population levels [80]. Ecological validity was further increased by including all students in the interventions and in the study. We will implement strategies to address potential selection bias, including collecting detailed demographic and baseline data and using inverse propensity score weighting using double robust estimators, if necessary.

Due to the absence of comparable previous research with the same study population, a priori sample size calculations were not conducted. This represents a limitation, as underpowered samples may fail to detect treatment effects, while overpowered samples may lead to unnecessary use of resources. Nonetheless, the existing literature indicates that significant treatment effects can be observed with smaller sample sizes in different populations [17,24]. Given the complexity of our analysis plan, which includes mixed model analyses and potential (double robust) inverse propensity score weighting, we believe it was justified to aim for a larger target sample size in this study.

We did not appoint an independent data monitoring committee, as this is a naturalistic trial without randomization, and it investigates interventions that focus on psychosocial well-being rather than alleviating psychiatric pathologies. Moreover, there were no clinical inclusion criteria, and as the participants were university students, most showed rather low levels of psychological distress. Therefore, we anticipated minimal risk of harm from the interventions, a decision that was also accepted by the ethical review board of the study. However, we recognize that the presence of such a committee could have enhanced the integrity and trustworthiness of the data collection process.

The study protocol was registered retrospectively. This lack of prospective registration may affect the perceived transparency and credibility of the work. However, in this protocol paper, we have aimed to enhance transparency by providing a comprehensive and detailed description of the study design and methodology before analyzing and publishing the results.

These results can be expected to be generalizable across Finnish higher education, but caution is needed when extending the results beyond the higher education context or to institutions in other countries. Future studies should investigate the implementation and effectiveness of these interventions in different contexts, including other educational institutions in Finland and internationally, to enhance the understanding of their applicability and impact.

The trial results will be published in peer-reviewed scientific journals and scientific conferences. We follow the authorship eligibility guidelines of the International Committee of Medical Journal Editors [81]. We do not intend to use professional writers.

If proven effective, the interventions outlined in this protocol could be implemented in universities nationwide and have a significant impact in improving university students’ well-being in both the short and long term, fostering positive mental health and supporting academic success and future career paths. These models may also be applicable in other settings, such as basic and secondary education, and in workplace environments. Ultimately, this research aims to support inclusion and equality in youth well-being services.