Mental health analytics: An innovative approach to understanding students’ wellbeing

Ensuring data sharing to support mental wellbeing is performed ethically and with the informed consent of students.

A partnership has been formed between higher education, charity and technology sectors and students’ unions, including the University of East London, Buckinghamshire New University, Civitas Learning, JISC and UUK to collaborate with Northumbria University on their mental health analytics project.

Approximately half of students with a mental health condition choose not to disclose it to their university (Thorley et al, 2017), with many students not accessing the support services that are available to them.

Data sharing within a university can provide a more comprehensive overview of individual students and potentially identify when their mental wellbeing may be at risk and they may require support. For example, being able to see that a student is not attending classes, showing declining grades, not paying their tuition fees, accessing learning resources in particular ways and not leaving their accommodation suggests a student in need of support.

The integration of data for a ‘whole view’ promotes more effective safeguarding through better co-ordinated multi-agency intervention – rather than multiple services and departments contacting the same individual independently. It also offers the potential for early intervention and response.

This project uses innovative integration of technology, advanced educational data analytics, student relationship management and student support to develop understanding of student mental health needs, and to offer early intervention through nudges signposting appropriate service support.

This includes highlighting key themes which will help university services understand when a student may be experiencing these thoughts to provide both early diagnosis and prevention.

Educational analytics is an emerging field whereby universities use data to identify students that share commonalities in data trends to students that have previously discontinued from studying (Foster and Francis 2019).

By identifying students with a greater likelihood of discontinuing, interventions are personalised to better signpost students towards sources of support. At Northumbria University, this intervention involves tailored ‘nudging’ emails. Students are given the option at enrolment each year to consent to the receipt of these tailored nudges.

Through this intervention and engagement strategy, a high number of students disclosed mental health difficulties they have been experiencing. This led to the hypothesis that the analytics system could be adapted to detect students that may be experiencing mental health difficulties.

However, unlike a retention model where data is available on the students who discontinue university, there are several challenges in building a mental health analytics platform as universities do not systematically collect data on the mental wellbeing of their students.

Therefore, from the outset it was important that students provided us with data on how they were feeling to develop a system more attuned to identifying students possibly experiencing low wellbeing.

While students ‘opt in’ for educational analytics it was unclear whether students felt the same about universities intervening to provide mental health support.

The challenge then was three fold:

1. Do students consent for their data to be collated from across the university for the purpose of developing mental health interventions?
2. Will students complete questionnaires about their mental wellbeing in order to support a more regular collation of student mental health data?
3. Do students consent to receiving tailored messages about seeking support for mental wellbeing?

The purpose of this transformative project is to identify actionable insights to deliver holistic approaches to student wellbeing and education. To address this challenge, we provided all students at enrolment the opportunity to:

• consent to their data being used as part of mental health analytics to receive tailored messages based on their likelihood of experiencing mental health difficulties. This included detail as to the nature of the project and likely implications of engagement in it
• complete a brief, five-item mental wellbeing questionnaire (WHO-5). This five-item questionnaire provides outcome data on which to build a predictive model and helps us identify students potentially at risk.

The WHO-5 questionnaire was delivered at a further two timepoints during the academic year 2020-21 in order to better understand the trajectory of mental wellbeing for students, and best ways of engaging with students.

The majority of students provided consent at enrolment to mental health analytics and completed the mental wellbeing questionnaire at the same time.

Of the 28,379 students completing the online enrolment, 61 per cent consented to take part in mental health analytics, with 56 per cent completing the mental wellbeing questionnaire. In contrast, 80 per cent consented to educational analytics.

The high consent rates are encouraging:

1. Consent rates for educational analytics at Northumbria University were ~60 per cent when initiated (now at ~80 per cent). Year-on-year increases in consent suggest that nudging as an intervention is perceived as beneficial to returning students.
2. When initiating educational analytics, several on-campus initiatives promoted uptake. This in-person engagement has not been possible during the coronavirus lockdowns.

Later in academic year – in the middle and end of semester two – students were invited to complete the WHO-5 questionnaire again, with completion rates of 12 per cent achieved for each timepoint.

Of those students engaging in this process, 33 per cent completed both questionnaires at the two timepoints – indicating that approximately 20 per cent of the student population completed at least one WHO-5 in the second semester.

Given the lower response rates to the WHO-5 through email, it stresses the importance of both embedding consent procedures at enrolment and not relying on email as a form of communication to reach most students.

However, the increased proportion of students with ‘Very low’ wellbeing scores at follow-up compared to enrolment (29 per cent versus 6 per cent) shows that reconnecting students at regular points during the academic year is an effective strategy to identify distressed students – further evidenced by spikes in support service self-referrals observed when simply asking about mental wellbeing (link to services sent with the questionnaire) (Figure 1).

Next steps include:

1. Establishing this system has shown disparities between student groups, such as Black students being the ethnic group most likely to consent to mental health analytics, but least likely to complete the WHO-5. There is a need to understand the differences between students groups in perceptions, trust and understanding of data systems set up to support mental wellbeing, and how these differences may be influenced by other intersectionalities such as disability, social deprivation and nationality.
2. The team is conducting interviews to evaluate students’ understanding and concerns about consent. This has provided key insights into support for the concept but also practical considerations that may influence decisions to engage:

   ‘If it can help future students have a better and more tailored positive experience of university then I'm keen to take part.’

   ‘If the analytics system does lead to students accessing the universities support services more, then there will need to be more funding for those provisions put into place.’

3. To review the most effective means through which students can be communicated at time points beyond enrolment. This work must be undertaken with the engagement of students in order to optimise communication channels for both students and the organisation.