Young adulthood is characterized by a multitude of fundamental transitions: from school to university, and then to employment. Mastering these transitions has been identified for many decades as a central developmental task of early adulthood (Arnett, 2000; Havighurst, 1972). The changes that university studies entail require significant adjustment and the development of coping strategies for the associated psychosocial stressors (Skinner et al., 2003). Time pressure due to deadlines and worries concerning exams are well-known risk factors for mental disorders among students, particularly depression and anxiety disorders, which are often associated with reduced motivation and academic performance (Duffy et al., 2020).

Recent studies report increasing stress and mental health issues among students (Heumann et al., 2024; Li et al., 2022; Tsiouris et al., 2023; Vidović et al., 2024). In an umbrella review of 62 meta-analysis with more than 8.7 million participants, Paiva et al. (2025) found relatively high prevalence rates for symptoms of mental disorders among university students, especially for depressive symptoms (mild: 35.41%, moderate: 24.54%, severe: 13.42%), anxiety (mild: 40.21%, moderate: 28.18%, severe: 16.78%), sleep disorders (mild: 41.09%, moderate: 23.3%, severe: 13.02%), and increased stress prevalence of 36.34%. In a survey of 24,533 students at 13 German universities, approximately 40–50% reported (almost) always or frequently feeling exhausted during their daily studies or experiencing burnout symptoms during the COVID-19 pandemic (Heinrichs et al., 2024). The changes in study conditions in many degree programs brought about by the COVID-19 pandemic (e.g., decrease in social contacts) may also be associated with increased rates of depressive symptoms especially in countries with higher income (Paiva et al., 2025), and could have a lasting impact on the mental health and qualifications of the next generation of academics, who are so urgently needed to build crisis-resilient societal structures.

However, according to international research, students with mental health needs show poor help-seeking activities, especially male students (Pei et al., 2024). At the same time, female undergraduate students have a higher likelihood of experiencing mental health issues (Kartikasari et al., 2025). These findings indicate that existing mental health services (MHS) fail to reach students comprehensively; instead, there is a demand for tailored interventions that are actually used, especially before the manifestation of symptoms requiring treatment.

Universities are central to student health promotion, encompassing both setting-based and behavioral prevention. The World Health Organization (WHO) has recognized the mismatch of mental health issues and actual uptake of MHS among university students several years ago (Cuijpers et al., 2019). Numerous universities have also acknowledged the urgent need to support students’ developmental tasks and have launched (more or less mandatory) nation-wide initiatives to promote students’ mental health around the world, e.g., in China (Yu et al., 2025), India (Jaisoorya et al., 2023) or Canada (Read et al., 2023). In contrast to voluntary programs and frameworks for promoting students’ mental health in countries like the United States (Rolin and Appelbaum, 2025) or Australia (Hutchesson et al., 2025; Luu, 2025), German policy makers have legally anchored this duty in the Prevention Act (§20 Social Code Book V), enshrining the funding of health-promoting programs on the expense of statutory health insurance companies for the vast majority of citizens, including university students (Hungerland et al., 2022). Universities typically offer or suggest varied psychological services, ranging from counseling centers to referrals to psychotherapeutic outpatient clinics with individual consultations. For instance, psychological counseling, available on-site and online, is a common and beneficial service (Cerolini et al., 2023; Pizzo et al., 2024), yet varies significantly by country, location, legal status, and institutional size (Franzoi et al., 2022). Although high-income countries typically offer a wide range of on-campus MHS, these resources are often underutilized, disjointed, or poorly coordinated, highlighting a key international challenge (Osborn et al., 2022).

Despite the high prevalence of mental illness among students (Heumann et al., 2024; Paiva et al., 2025), university MHS are underutilized, especially among vulnerable student groups (Hyseni Duraku et al., 2024). A systematic review (Zhao et al., 2025) found that only 28% of students with mental health issues actively sought help, though 41% intended to.

Barriers to seeking help are both individual and institutional. At the individual level, key obstacles include self-reliance, lack of time, limited awareness of resources, under-recognition of early symptoms, and the belief that they do not need formal support (Zhao et al., 2025; Ebert et al., 2019; Krümmel et al., 2023; MacDonald et al., 2022). At the institutional level, fragmented responsibilities, regulatory obstacles and limited resources further hinder comprehensive psychosocial support (Hyseni Duraku et al., 2024; Priestley et al., 2022; Roy et al., 2025).

To extend on-campus MSH, such as stress management, mindfulness and relaxation courses, flexible low-threshold digital MHS have also been increasingly developed, showing efficacy in reducing stress and depressive symptoms among university students (Harrer et al., 2018; Harrer et al., 2021; Matos Fialho et al., 2025). The initial barrier, though, is students recognizing their need for support and the drivers for utilization for long-term preventative effects (Krümmel et al., 2023). Subsequently, they must actively register for these MHS, a process particularly challenging for first-year and international students unfamiliar with the diverse range of university support providers (Schweighart et al., 2024).

Additionally, despite the initial optimism surrounding pilot programs, the sustainable integration of digital MHS into university structures remains limited (D’Adamo et al., 2023), and students lack information regarding trustworthy digital MHS (Montagni et al., 2020; Vomhof et al., 2024). Furthermore, universities often lack financial and personnel resources for the permanent implementation of quality-approved digital MHS, and costs for implementation as well as maintenance of digital MHS for university students are seldom reported (Taylor et al., 2024). Integrating mental health promotion and education on self-help digital MHS into curricula (e.g., Car et al., 2025; Sahan et al., 2024; Till et al., 2024) is promising, but oftentimes restricted to specific subjects (e.g., medicine). Further established and implementable solutions exist to improve service awareness and uptake, such as offering app-based recommender systems, including stress screening and feedback tools, such as “TrackYourStress” (Pryss et al., 2019), which provide individualized service recommendations and facilitate registration and initial engagement.

From a technical perspective, the platform will fully leverage state-of-the-art technologies to maximize user acceptance, while providing transparency by adhering to reporting guidelines for HRS (De Croon et al., 2021). Beyond the already mentioned participatory development approach, we identify the following technical features as essential, as shown in Table 1.

To support these capabilities, modern cross-platform architecture will be implemented, consisting of a web application (fully responsive for mobile devices) and native mobile apps for Android and iOS that can be distributed via official stores. The architecture of the sensor framework is detailed elsewhere (Karthan et al., 2022). Ideally, the apps should also connect with relevant external stakeholders (e.g., health insurers, counseling centers) to ensure real-time access to updated content.

In order to ensure a high privacy and security standard, we intend to clearly separate Identity Data (IDAT) from Medical Data (MDAT), which is a common approach in medical informatics (Lablans et al., 2015; Spitzer et al., 2009). Login information, including the e-mail address containing students’ names and profile data, will be stored within the IDAT database, while all collected information and recommendations are stored in a separate MDAT database. Depending on the system, there may be two different databases (i.e., like two different relational databases), two different files (i.e., like two different SQLite databases), or one large database with two different schemas (i.e., PostgreSQL). Most importantly, this approach guarantees that no identifying information is stored within the MDAT database. Additionally, both databases should be encrypted by default (with different keys).

To enhance students’ acceptance, a mixed-methods, co-creation approach will be applied. In a first step, interviews with experts from respective domains are conducted to assess basic requirements of the platform (e.g., instruments, features of the recommender system, etc.). Next, a comprehensive user-interface will be developed and evaluated with target end users to meet their demands. Feedback will be collected via an online survey and incorporated in the apps via rapid-development cycles (e.g., using the System Usability Scale; Bangor et al., 2008; Hyzy et al., 2022).

Initially, the recommender system balancing personalization and privacy will mainly process user-entered data on needs (e.g., Ecological Momentary Assessment; EMA) and stated preferences for hybrid filtering of suitable MHS (using content- and knowledge-based, see Table 2). The acceptance of collaborative filtering and more sensitive data collection required for context-aware filtering will be investigated using participatory research approaches prior to possible implementation.

Before using the app, students will provide informed consent, clearly understanding how and why their data is used and making an informed decision about the personalization-privacy trade-off. Our recommender system will initially use a hybrid approach (content- and knowledge-based filtering), which is safe and explainable, matching students’ EMA responses and stated preferences with relevant MHS. While this can become repetitive, context-aware filtering can provide just-in-time suggestions if students agree to share more sensitive data (e.g., sensor data). We recognize that collaborative filtering offers novel, peer-endorsed suggestions but requires a large dataset and risks exacerbating bias (Stinson, 2022). Finally, we propose federated learning (FL) for privacy-preserving model training directly on user devices. FL will improve cross-institutional generalizability, enhance trust in personalized MHS suggestions across multiple universities (using the same EMA instruments), and enable larger-scale performance, potentially in an international context. Nonetheless, FL approaches are still at an early stage (Grataloup and Kurpicz-Briki, 2024).

We plan to evaluate the system through a controlled field study (e.g., A/B trial) to assess its effectiveness regarding changes in students’ mental health outcomes based on the specific MHS it recommends. In addition, we will employ a mixed-methods adoption assessment, combining usage analytics with online surveys and qualitative interviews to understand user acceptance, perceived usefulness, and factors influencing sustained engagement as well as practical applicability (see Table 1). A concrete study design must be discussed with stakeholders from different domains in order to gain broader insights.

From a content perspective, the platform concept includes the following modules (see Figure 1).

Module 1: “I can learn more about myself and my health.”

The screening tool is based on EMA (Shiffman et al., 2008) that can be augmented with a mobile crowdsensing solution (Pryss et al., 2018). Data gathered through crowdsensing can be used to identify university-specific stress scenarios in larger student groups, such as exam situations and the possibility to provide feedback to students (Pryss et al., 2019; Pryss et al., 2019). Students have access to various screening tools (e.g., subjective stress levels, dealing with test anxiety, or time management problems) and receive individualized feedback on their mental health status. Through continuous use of such a digital platform via a mobile app in everyday life, students are provided with on-time feedback on their stress levels and stress-inducing situations in their study routines. The screening can be complemented with psychoeducation and brief exercises. Such immediate support is important, as students experiencing acute stress (e.g., due to exams) may prioritize targeted short-term relief on long-term preventive measures (Krümmel et al., 2023).

Module 2: “I can use suitable health services at my university.”

Based on the results of the screening tool (see Module 1), an individualized recommendation of support services can be made at the respective local university as a trustworthy information source (Vomhof et al., 2024; Braun et al., 2023). This module acts as a kind of guide, selecting relevant MHS from a stored database and presenting demand-tailored suggestions to the students (e.g., sorted by suitability or personal relevance). The platform may contain a uniform registration form for the federated authentication for respective events or services, so that students can register directly. This eliminates several access barriers, such as different registration forms or information channels. Module 2 aims to facilitate access to existing MHS through comprehensive information and low-threshold registration options (Apolinário-Hagen et al., 2023).

Module 3: “I can participate in a digital health service.”

To increase the reach of health promotion, many universities already provide their students free access to a range of digital mental health resources, such as “TrackYourStress” (Pryss et al., 2019; O’Rourke et al., 2022). Some MHS have been developed in-house by universities. Others are offered free of charge to students in cooperation with partners, such as health insurance companies. Besides on-site interventions, many students are unaware of available digital MHS (Schweighart et al., 2024; Apolinário-Hagen et al., 2021). Likewise, the overly wide range of registration options for digital MHS represents a key barrier to access and uptake among students (Dederichs et al., 2021) as well as perceived lack of personalization in digital MHS (Riboldi et al., 2024). The goal of the “ConnectedHealth@University” platform is to provide personalized guidance on digital MHS, such as “StudiCare Stress” (Harrer et al., 2021). This includes direct download and registration options for digital MHS to enable low-threshold access (see Supplementary Figure S1).

Module 4: “I can evaluate the health services offered by my university.”

Although methods for measuring user satisfaction and acceptance of MHS are well-described, systematic collection of feedback data is rarely used at universities, especially by student counseling centers that often have limited personal resources as well as no access to suitable digital documentation systems (Farahani et al., 2023). In addition, there is scarce systematic research on the actual utilization rates of these MHS for students, with globally largely varying rates (Osborn et al., 2022).

The platform “ConnectedHealth@University” incorporates a module to establish consistent and continuous feedback data on MHS and a satisfaction monitoring system for existing MHS. To achieve this, the platform regularly sends notifications, asking students to rate respective programs. The feedback results may be shared with relevant stakeholders at each university and through networks. These reports are anonymized and provided in compliance with data protection regulations such as GDPR for European settings, using a user-friendly dashboard solution.

Ethical challenges, including fairness and equity, must be considered throughout the development of recommender systems, especially in FL (Grataloup and Kurpicz-Briki, 2024). Self-selection bias is a common challenge, as current digital MHS research predominantly relies on female and White participants (Taylor et al., 2024). Students who participate are often those already informed about support, and thus may not represent the primary target group. To avoid this, recruitment strategies for co-creation must be adapted (Hasanzadeh et al., 2025). We will gather historical data from diverse student subgroups using the established “TrackYourStress” app (Pryss et al., 2019; O’Rourke et al., 2022) that applies validated EMA measures. This will be complemented by larger representative survey data for the platform’s “cold start,” as recommended (Hasanzadeh et al., 2025; Stinson, 2022). Our project will start with at least four different universities across Germany.

Personalization based on sociodemographic characteristics such as ethnicity will be avoided to prevent oversimplification and prejudice perpetuation (Taylor et al., 2024). Instead, personalization will use assessed needs (via EMA) and stated preferences. The acceptance of further features (context-aware filtering, preferred authentication) will be explored in formative qualitative research (e.g., focus group discussions). Since EMA and other health data are sensitive under GDPR (Ienca and Malgieri, 2022; Zhang et al., 2025), we will prioritize transparency and user control. Students will be fully informed about data use with a clear opt-out option. Because the platform targets health promotion and early intervention, not acute crises, a minimal safety protocol will be implemented, automatically triggered by elevated stress scores to provide immediate, generic, expert/rule-based information on regular support services. The platform will neither include questions about suicidal ideation or clinical screening, nor provide direct access to healthcare. Students will be informed that the system is not a substitute for therapist contact or suitable for severe issues.

Sustainable platform implementation requires context-sensitive, early involvement of key stakeholders (Cross et al., 2025), such as on-campus psychological counseling centers as gatekeepers, and practical guidance for university staff on updating MHS in the platform. Staff buy-in is thus crucial, necessitating measures like participatory approaches to increase commitment, like communicating benefits (e.g., reduced workload for general advice), and providing training and support. In addition, the platform must be integrated into existing IT infrastructure, including digital support systems of counseling centers. Ongoing maintenance should be ensured through the platform’s design (reliable, safe personalized recommendations via FL; shift to passive data collection/tracking; reduced effort for students and staff) and long-term implementation strategies, including funding, workflow integration and training (Berardi et al., 2024; Löchner et al., 2025; Nair et al., 2024). Both institution-wide and cross-university governance must be considered.

German universities benefit from the Prevention Act (Hungerland et al., 2022) for establishing cross-institutional MHS recommender systems. A realistic approach involves using competitive funding schemes for initial piloting and validation, while engaging health insurance representatives early to secure future implementation.