The research, published in the Journal of Autism and Developmental Disorders, was led by Hatice Gundeslioglu of the University of Warwick together with colleagues at the University of Birmingham and Monash University. The team followed UK undergraduates across two waves of an online survey spaced six months apart, comparing autistic students, identified through self-reported clinical diagnosis, with their non-autistic peers. Their central question was deceptively simple: do anxiety, depression and suicidality feed into one another over time, and does being autistic, or being a particular gender, change those dynamics?

To answer it, the researchers used a statistical technique called a cross-lagged panel model, or CLPM. Unlike a simple correlation, which cannot tell you whether A drives B or B drives A, a cross-lagged design measures the same variables at two time points and asks whether each construct predicts change in the others after accounting for its own stability. In this study, the first wave recruited 747 undergraduates between November 2022 and June 2023; six months later, 209 participants completed the follow-up, and 179 provided complete data for the final models. Mental health was measured with well-validated instruments, including the GAD-7 for anxiety and the PHQ-9 for depression, alongside items on self-harm and suicidal ideation.

An exploratory factor analysis distilled the questionnaire items into three distinct dimensions: anxiety and worry, behavioural symptoms of depression and anxiety, and suicidality. The resulting models showed an exceptionally good statistical fit, with fit indices at or near their ideal values. That matters because it gives confidence that the patterns the researchers report are not artefacts of a poorly specified model, but reflect genuine structure in the data.

The headline finding is directional. Suicidality at the first wave significantly predicted increases in both anxiety and worry and in behavioural symptoms of depression and anxiety at the second wave, with standardised coefficients of 0.21 and 0.25 respectively. These are small to moderate effects, but in a longitudinal design even modest effects can accumulate into clinically meaningful change. In other words, a student who reported suicidal thoughts or self-harm at baseline was more likely to be more anxious and more depressed six months later, even after the researchers statistically accounted for how stable those problems already were.

Equally striking was the sheer stability of mental health across the six-month window. The autoregressive paths, which capture how well a construct predicts itself over time, ranged from roughly 0.70 to 0.80 across the three outcomes. Values in that range indicate high but not perfect stability: most of what determines a student’s mental health at the follow-up is simply where they stood six months earlier. Students who were struggling at the start of the study were, by and large, still struggling at the end, a pattern consistent with earlier work showing that psychological distress among undergraduates tends to persist across an academic year rather than resolve on its own.

Where does autism fit in? At baseline, autistic undergraduates reported significantly higher levels of anxiety, depression and suicidality than their non-autistic peers, confirming what previous cross-sectional research had suggested. Yet, perhaps counterintuitively, an autism diagnosis did not predict change in mental health over the six-month period. Once baseline levels were taken into account, autistic and non-autistic students followed similar trajectories. The authors suggest several possible explanations: some autistic students may have received targeted support or developed effective coping strategies; the autistic group is highly heterogeneous in support needs and circumstances; or six months may simply be too short a window to capture divergence that emerges over years.

Gender told a similar story of null effects. Neither gender nor autism diagnosis was significantly associated with mental health outcomes at follow-up in the cross-lagged models. That finding runs against some prior studies reporting that female students experience more stress, anxiety and depression during university, and against evidence that autistic females face heightened suicide risk. The authors caution that their sample was modest, skewed by attrition, and recruited largely through online communities, which may have masked underlying gender differences. Notably, they could not test whether autism and gender interact, for example whether autistic female students fare worse over time than autistic male students, because the follow-up sample of 179 fell well short of the 300 participants their pre-registered analysis plan called for to support such complex models.

The study’s limitations deserve honest attention. Attrition was substantial: only about 28 percent of the original wave-one sample completed the follow-up, and those retained differed from those lost, with autistic students retained at higher rates. Missing data were not completely at random, forcing the researchers to rely on complete-case analysis, which reduces statistical power and may limit generalisability. Autism status rested on self-report of a clinical diagnosis, and eleven participants reported receiving their diagnosis during the study interval itself. The six-month follow-up is also short by longitudinal standards, and the authors recommend future studies with at least three waves and longer intervals to capture change more reliably.

Even so, the practical implications are hard to ignore. If suicidality predicts subsequent increases in anxiety and depression, then screening for suicidal thoughts and self-harm, and intervening early, could blunt the development of broader mental health problems in the student population. The stability of symptoms over time argues for regular, long-term monitoring rather than one-off assessments, so that universities can identify students whose difficulties persist and intervene before they deepen. And although no overall gender differences emerged in this analysis, the wave-one data from the same cohort showed that autistic female students reported elevated suicidality, leading the authors to suggest that university well-being services should pay particular attention to this subgroup. For a generation of students, autistic and non-autistic alike, the message from this research is clear: suicidal distress is not just a symptom to be managed after the fact, but a signal that deserves immediate, sustained attention.

Subject of Research: Longitudinal bidirectional relationships between suicidality, anxiety and depression among autistic and non-autistic UK undergraduates

Article Title: A Cross-Lagged Panel Model of Mental Health Amongst UK Undergraduates: The Role of Autism and Gender

Article References: Gundeslioglu, H., Gray, K. M., Thompson, P. A., & Langdon, P. E. (2026). A Cross-Lagged Panel Model of Mental Health Amongst UK Undergraduates: The Role of Autism and Gender. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07541-8

Image Credits: AI Generated

DOI: 10.1007/s10803-026-07541-8

Keywords: autism, undergraduates, mental health, suicidality, anxiety, depression, cross-lagged panel model, longitudinal study, gender, UK universities, GAD-7, PHQ-9