The study followed the Consolidated Criteria for Reporting Qualitative Research (COREQ) checklist (Checklist 1).

This study used a qualitative methodology to explore a “how” research question, focusing on the depth and complexity of an underresearched topic. Guided by a relativist ontology, subjectivist epistemology, and an interpretivist paradigm [31], our approach is well-suited to research in psychology, mental health, and education, where lived experience and sociocultural context are central [32,33]. A grounded theory methodology supported the inductive development of insights from participants’ perspectives [34]. Semistructured interviews were used to gather rich, nuanced data, aligned with the study’s exploratory aims [35].

Participants were identified through student networks, social media platforms (Instagram [Meta], X, and LinkedIn), and snowball sampling, an approach well-suited for exploring sensitive topics and effective in reaching diverse student populations, including international students [32,33]. Interested participants were encouraged to email the researcher to express their interest. Participants were eligible to take part if they were students (aged 18‐25 y) enrolled in an undergraduate or postgraduate course at a university in London, fluent English speakers, and willing to provide informed written consent. The age range represented students enrolled in undergraduate or postgraduate taught programs, which typically involve more structured online learning. Efforts were made to include diversity in gender, ethnicity, sexual orientation, degree, and year of study. All individuals who expressed interest and were contacted by the researcher met the eligibility criteria; none declined to participate or dropped out of the study. Participation was voluntary, and as this study was part of a master’s thesis, no compensation was provided.

Participants’ demographic data were collected at recruitment (Table 1). Semistructured interviews were conducted by the lead researcher (AF) via Microsoft Teams between June 6 and July 27, 2023. Open-ended questions explored participants’ knowledge, attitudes, and concerns regarding the use of learning analytics to support university students’ mental health and well-being. The semistructured interview questions were informed by a review of literature on student mental health, learning analytics, and help-seeking in higher education. They were also guided by the technology acceptance model [36] to explore perceptions of learning analytics and self-determination theory [37] to consider factors influencing student motivation and engagement. Furthermore, 3 university students informed the topic guide (refer to Patient and Public Involvement section). It consisted of five sections: (1) introductory questions to help participants feel comfortable and at ease, (2) existing mental health support at university, (3) use of learning analytics to support university students’ mental health and well-being, (4) facilitators and barriers to learning analytics adoption for mental health and well-being purposes, and (5) integration of learning analytics within existing university mental health support services (Appendix S1 in Multimedia Appendix 1). All interviews were audio- and video-recorded, then transcribed verbatim. No one else was present during the interview except for the participants and the researcher. Transcripts were accessible to participants through Microsoft Teams. Field notes were taken during and after the interview and were considered when coding the transcripts.

We advertised for individuals with lived experience (ie, university students whose data were being used by the university) through the Imperial College Union Mental Health Network Instagram account to join our team as advisors and inform the project. As a result, 3 students, aged 21‐26 years (2 female and 1 male), from the Faculty of Natural Sciences and the Faculty of Medicine at Imperial College London expressed interest and joined the team to form a student advisory group (SAG). The SAG included 1 postgraduate and 2 undergraduate students and informed 3 key stages of the research cycle [38]: designing and managing research, carrying out the research study, and dissemination.

We held a 2-hour online meeting (Microsoft Teams) on April 7, 2023, where the SAG informed the research objectives and accessibility and appropriateness of project documentation (interview topic guide, participant information sheet, consent form, and protocol). The following changes were made to the topic guide: language was simplified (ie, a lay definition of learning analytics was agreed upon), and a question was added regarding the impact of learning analytics on academic outcomes. The final topic guide was piloted with the SAG in 3 mock interviews. After providing all positive feedback, no subsequent changes were made to the topic guide. Following the generation of initial codes, a coding framework was cocreated with the SAG during an additional 2-hour online meeting (Microsoft Teams) on July 14, 2023. We held a final meeting on August 2, 2023, where the SAG and the 2 researchers (AF and MO) refined and consolidated themes. Several themes were revised, consolidated, and more effectively linked through student advisor involvement (Figure S1 in Multimedia Appendix 1). The SAG was not compensated for their time, as no budget was associated with this study, being a master’s thesis. Patient and public involvement (PPI) was reported using the Guidance for Reporting Involvement of Patients and the Public, Version 2-short version [39] (Table S1 in Multimedia Appendix 1).

An inductive and deductive method to Braun and Clarke’s 6-stage thematic analysis guided our analysis [40]. We used a data-driven approach to theme development while also drawing from the researcher’s previous knowledge of learning analytics. Transcripts were imported into Microsoft Word. After transcription, the lead researcher (AF) familiarized herself with the data by reading each transcript multiple times. All transcripts were manually coded in Microsoft Word, and relevant codes were compiled. Manual coding was chosen as it supported reflexivity, the development of the lead researcher’s skills, and enabled deep immersion in the data [41-43]. Furthermore, 2 independent researchers (AF and MO) compared codes from 2 transcripts during a 2-hour online meeting on Microsoft Teams. Data triangulation helped ensure the reliability and validity of the findings by identifying potential individual researcher biases and confirming the consistency of information. The coding framework was applied to code the remaining transcripts. Interviews were conducted until data saturation was reached, assessed through concurrent data analysis, so that no new codes emerged, and responses across participants consistently reinforced existing patterns. AF then organized codes into potential themes and subthemes and gathered all data relevant to each of these. Some themes were combined to avoid redundancy, while others were broken down into separate themes and judged for internal homogeneity and external heterogeneity. The thematic map was finalized with the SAG.

All procedures involving human participants were approved by the Head of Department of the School of Public Health and the Research Governance and Integrity Team at Imperial College London (Imperial College Research Ethics Committee reference 6604462). The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.

Participants provided informed consent by signing and returning a consent form via email. A copy co-signed by both the participant and the researcher was emailed back to the participant and stored securely on Imperial College London encrypted drives accessible only to the study team. Participants could withdraw at any time before the start of data analysis. Audio and video recordings were deleted after transcription, and study-specific ID numbers were used to maintain participants’ anonymity. The principal investigator ensured confidentiality and General Data Protection Regulation compliance for health and care research. Participation was voluntary and uncompensated, as the study formed part of a master’s thesis.

In total, 15 participants took part in the study (10 female, 5 male), with a mean age of 21.47 (SD 1.60, range 19‐24) years (Table 1). Students from 4 London universities participated in the interview (Table S2 in Multimedia Appendix 1). Most participants were undergraduate students (11/15, 73%). All students were enrolled in a science, technology, engineering, and mathematics (STEM) degree. Interviews lasted between 19 and 34 minutes.

There were three themes: (1) potential of learning analytics for mental health and well-being innovation, (2) student involvement in decision-making regarding learning analytics, and (3) integration of learning analytics with existing support (Figure 1).

To the authors’ knowledge, this is the first study to explore students’ perceptions of learning analytics to support students’ mental health and well-being. There were three main themes: (1) potential of learning analytics for mental health and well-being innovation, (2) student involvement in decision-making regarding learning analytics, and (3) integration of learning analytics with existing support.

All participants supported the use of learning analytics as a tool to support students’ mental health and well-being. The majority of participants felt that learning analytics could enhance access to and delivery of support services, advocating for an evidence-based and student-centered approach, in line with existing qualitative literature and policy recommendations on improving student mental health support [35-38]. However, some students also recognized potential challenges in using learning analytics to detect and intervene in case of student mental health deterioration. Overall, our findings show that students’ positive attitudes and acceptability of learning analytics hold promise for implementation within current university support systems. Learning analytics may help enhance coordination between academic and mental health services, supporting UK policy recommendations to adopt a whole-university approach to mental health [4].

Despite an initial lack of awareness of learning analytics, all students recognized its potential beyond the learning context to support mental health and well-being. This aligns with qualitative literature that reports students’ generally positive attitudes toward learning analytics [17,25] and digital tools for mental health monitoring and early intervention in young people [44-46]. Most participants believed that learning analytics could provide objective and reliable evidence for monitoring students’ mental health. They also welcomed its potential to inform early interventions and integrate with existing university support systems. However, they also recognized the potential challenges of using learning analytics for mental health purposes and stressed the need to be involved in decision-making about learning analytics development and implementation.

Several barriers and facilitators were identified for adopting learning analytics to support students’ mental health and well-being. Some students raised concerns regarding the reliability of learning data and identified several ethical issues, in line with other qualitative studies [14,17,18,25]. Students worried that data inaccuracy could lead to delivering inappropriate interventions, highlighting the need for a context-specific understanding of students’ behavior. Students also reported the importance of balancing the need to share data for effective interventions with concerns about privacy breaches and ensuring informed consent, reflecting findings from other studies [20,26,47-49] and aligning with proposed recommendations for ethics and privacy in learning analytics [50,51]. Participants agreed these risks might subside if universities were more transparent about using learning analytics and abided by good practice guidelines [14,15,28,30].

Overall, students’ ethical concerns underscore the need to adopt a coproduced approach in the development, implementation, and governance of learning analytics that strategically incorporates the student voice and aligns with the principles of PPI [15,16]. Meaningfully engaging students as active stakeholders, beyond their role as data subjects, is essential to fostering trust and ensuring that learning analytics systems are transparent, ethical, and responsive to student needs [27-30]. Regarding facilitators for learning analytics adoption, students stressed the need for incentives to engage with learning analytics, as well as staff inclusion and training, in line with other studies [19,21,23,52]. Our results support previous research [17,26], showing that learning analytics can be used to assess students’ learning methods and adjust teaching approaches and curricula to better support their mental health and academic needs. AI-driven computational approaches, including machine learning, natural language processing, and rule-based systems, were viewed positively. These methods were considered valuable additions to learning analytics systems, offering predictive capabilities and data-informed recommendations to enhance student learning and support students’ mental health when implemented with appropriate student involvement, as highlighted in a recent systematic review [53]. This aligns with qualitative research reporting young people’s acceptance of AI [54]. Therefore, our study suggests that students’ positive views on learning analytics support the potential for integrating learning analytics with existing academic and mental health support, as per UK policy recommendations [4].

The researcher’s perspective was informed by her lived experience as a student with mental health issues (Appendix S2 in Multimedia Appendix 1). This insider positionality facilitated access, built rapport, and enabled the development of informed, contextually relevant questions. However, it also introduced potential for unconscious bias, including overidentification with participants and possible influence from pre-existing views on university mental health provision. Furthermore, the researcher’s involvement in mental health advocacy within her institution may have led participants to have some awareness of her motivations for conducting the research.

To address these biases, a reflexive approach was used throughout the study to critically examine how personal assumptions, values, and previous knowledge might influence design, data collection, and interpretation [55,56]. Attention to the insider-outsider dynamic, reflexive journaling, PPI, and data triangulation with other researchers was used to ensure transparency and analytical rigor, grounding the findings in participants’ perspectives rather than the researcher’s personal views.

There are several strengths to this study. A total of 3 people with lived experience and interest in mental health research informed the design, analysis, and interpretation research stages [57]. Students had personal knowledge and experience of mental health support at university, which made the research more relevant, focused, and accessible.

There are also study limitations. Despite efforts to maximize variation, participants’ backgrounds were not diverse in terms of gender and sexual orientation. Our sample was predominantly female, and we did not include individuals who did not identify with the gender assigned at birth. We also did not explore differences in sexual orientation, and our sample was predominantly heterosexual. LGBTQ+ students are particularly vulnerable to mental health problems [58], and their views would enrich the field. While the sample included both home and international students from a range of cultural and ethnic backgrounds, the study did not explicitly explore how cultural factors shape mental health experiences or help-seeking behaviors [59]. Furthermore, although participants came from multiple London universities, they were all STEM students, so our results may not be generalizable to other universities in the United Kingdom or students from other educational backgrounds. Future research could incorporate focus groups to explore the topic in a more naturalistic setting, thereby validating findings and minimizing researcher bias. Larger studies are also needed to enhance generalizability. Quantitative approaches, such as exploratory factor analysis, could be used to examine broader patterns based on the themes identified in this study.

This research focused on students’ perceptions of learning analytics to support students’ mental health and well-being. To ensure the successful adoption of learning analytics systems to support students’ mental health and well-being in higher education in the future, there is now a need to explore the perceptions of all other stakeholders, including academic staff, counseling and disability services, and IT services [60]. Our qualitative findings could also be complemented with a national survey across all UK universities and internationally to capture broader perspectives on the interview themes and explore relationships between the data. As learning analytics systems are implemented, future research should investigate the feasibility, acceptability, and effectiveness of these systems in improving students’ mental health and well-being.

There are also several practical and policy implications to this study. First, students’ general positive perceptions around the use and integration of learning analytics for mental health reflect the need to align learning analytics practices with mental health strategic priorities as part of a whole-university and data-driven approach to mental health [4]. Second, students’ ethical concerns regarding learning analytics adoption, particularly around transparency, consent, and privacy, highlight the importance of involving students in their development and implementation, echoing previous policy calls [15,16,27-30]. Universities should, therefore, develop clear institutional policies and protocols coproduced with students regarding the development and implementation of learning analytics. Third, institutional learning analytics policies should align with national higher-education–wide guidelines to ensure sector-wide consistency. A coordinated government strategy on student mental health is needed to align institutional practices with broader policy goals. Ultimately, while learning analytics offer strong potential to enhance student support and service delivery, their success relies on meaningful student involvement, institutional accountability, and coherent policy at both institutional and national levels.

This was the first study to explore students’ perceptions of using learning analytics to support students’ mental health and well-being. Our findings suggest that students perceive value in utilizing learning analytics to enhance mental health and well-being in higher education, as well as their integration with existing university services, which holds significant importance for UK policy. However, considering the ethical concerns raised by students, this study also highlights the need to inform students about learning analytics–based systems being developed or implemented at their institution, echoing previous calls for student engagement in the decision-making process regarding learning analytics. Ensuring learning analytics systems meet students’ needs and expectations is essential to avoid resistance to learning analytics adoption and empower students to make decisions regarding their welfare and educational experience. Overall, learning analytics holds promises for its integration within university support, but not without the involvement of students.