SADIE: a longitudinal Survey on Anxiety, Depression, Internalising and Externalising symptoms in Italian university students – a study protocol

Abstract

Introduction

University students experience escalating academic, social and financial stress, resulting in high prevalence of anxiety (18%–40%), depression (12–35%) and burnout (20–30%). The COVID-19 pandemic intensified these issues, but most studies report symptom remission post-lockdown, highlighting the need for long-term prospective data on symptom trajectories and risk factors. The Survey on Anxiety, Depression, Internalising and Externalising symptoms (SADIE) study aims to (1) estimate the prevalence and incidence of anxiety, depression, burnout and suicidal ideation in Italian university students over 36 months and (2) identify modifiable psychological and behavioural predictors—such as multidimensional perfectionism, academic stress and substance use—associated with adverse symptom trajectories.

Methods and analyses

We will recruit 5000 undergraduates and postgraduates (aged ≥18 years, fluent in Italian or English) from 14 Italian universities. Participants will complete online assessments via REDCap at baseline, 6-month, 18-month and 36-month follow-ups. Measures include demographic survey, Maslach Burnout Inventory-2, Beck Depression Inventory–II, Generalized Anxiety Disorder–7, WHO-5 Well-Being Index, Alcohol Use Disorders Identification Test-3, Cigarette Dependence Scale-5, Columbia Suicide Severity Rating Scale and Hewitt and Flett Multidimensional Perfectionism Scale. Mixed-effects models (R/lme4) and growth-mixture models will be used to evaluate symptom trajectories, while logistic regression will identify predictors of adverse outcomes. Sensitivity analyses will address potential self-selection and attrition bias.

Ethics and dissemination

The study has received approval from the Ethics Committee of the University of Padova (Protocol No. 235-d). Participants will provide electronic informed consent and may withdraw at any time. Risk management includes automated alerts and links to local mental health services for high-risk responses. Study findings will be disseminated through peer-reviewed publications, conference presentations and summary reports to participating institutions; deidentified data may be made available on reasonable request in line with institutional and data protection regulations.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The study uses a prospective, multicentre longitudinal design, tracking university students across 14 Italian institutions at four distinct time points over 36 months.

• Core outcomes and key psychological and behavioural predictors are measured using standardised, widely used psychometric instruments, enhancing reliability and comparability across sites.

• Planned analyses employ generalised linear mixed-effects models and growth-mixture models to capture heterogeneous symptom trajectories and identify latent high-risk subgroups.

• Recruitment through open, voluntary participation and convenience sampling may introduce self-selection bias and limit the generalisability of findings to the wider Italian student population.

• Long-term online follow-up over 36 months carries a substantial risk of attrition, which may affect precision of trajectory estimates despite planned use of weighting and full-information maximum likelihood methods.

Introduction

Background and rationale

University life is a pivotal developmental period marked by newfound independence, intensified academic demands and complex social dynamics. While this transition offers opportunities for personal and intellectual growth, it also exposes students to stressors that can compromise psychological well-being. At a population level, recent estimates from the Global Burden of Disease Study 2023 indicate that mental disorders affected approximately 1.17 billion people worldwide in 2023 and are now among the top five causes of global disability-adjusted life-years and the leading cause of years lived with disability.¹ Rigorous coursework, performance pressures, financial burdens and the challenge of establishing and maintaining social support networks contribute to a mounting load of academic and interpersonal stress. In line with prior evidence, university students report elevated levels of anxiety and depressive symptoms; in some studies, these levels are even higher than those observed in non-student peers.² This evidence comes from both single-institution surveys³ and systematic reviews focusing on the student population.⁴ Consistently, a recent comprehensive review in The Lancet emphasises that depression is a common illness affecting people in every society worldwide and represents a major contributor to disability and premature mortality.⁵ These patterns have concrete implications for functioning, including academic performance and persistence, and may set the stage for suicidal ideation in a vulnerable subset.^{6 7}

Recent evidence from large-scale Italian cohorts further underscores this need. In particular, Meda et al used machine-learning approaches to identify predictors of severe mental health problems among university students, revealing that traditional demographic and academic variables accounted for only a limited portion of the variance in symptom severity.⁸ These findings highlight the importance of incorporating multidimensional psychological and behavioural constructs, such as perfectionism, stress and coping styles, into longitudinal designs aimed at understanding mental-health risk trajectories in higher education.

Burnout in students

Burnout, characterised by emotional exhaustion, cynicism and a sense of reduced personal accomplishment, has emerged as a distinct phenomenon in student populations. Estimates from undergraduate student population indicate that 24.1% of undergraduates experience burnout, which in turn correlates with poorer academic performance, decreased retention and diminished quality of life.⁹ The interplay between burnout and internalising symptoms (eg, anxiety and depression) is bidirectional: chronic stress and exhaustion can precipitate affective disorders, while anxiety and depressive symptoms may exacerbate perceptions of academic overload. In high-demand academic programmes such as medicine, elevated perceived stress and depressive symptoms have been documented among students¹⁰, often accompanied by signs of emotional exhaustion and suicidal ideation. These findings suggest that the same interactions between academic burnout and internalising symptoms observed elsewhere may also be present in high-pressure Italian university contexts.

Impact of the COVID-19 pandemic

The advent of the COVID-19 pandemic introduced an unprecedented global stressor, disrupting traditional learning environments and exacerbating feelings of isolation. Longitudinal studies conducted in Italy have shown acute increases in anxiety and depressive symptoms during lockdown periods, followed by partial or complete remission with the resumption of in-person classes.¹¹ Moreover, cross-sectional data confirm that during the pandemic, university students reported greater psychological distress than other population groups.¹² However, individual variability in recovery pathways underscores the importance of pre-existing vulnerability factors—such as maladaptive perfectionism, repetitive negative thinking and low social support—in shaping long-term psychological outcomes.¹³ Convergent sources from Italian university counselling services during the pandemic corroborate both the scale and the profile of presenting problems—predominantly anxiety and depressive symptoms—and describe rapid expansions of remote support with uneven regional coverage and access.¹⁴

The Italian context: mental health and suicidality among university students

Zooming in on Italy, recent syntheses and service-level audits extend the picture beyond headline prevalence and lockdown spikes by clarifying who is most affected, where capacity is strained and which levers appear modifiable. A 2025 national meta-analysis (~23 000 students) confirms a high aggregate burden but, crucially, maps clusters of vulnerability (academic overload, financial strain, perfectionistic concerns and sleep disturbance) and protection (social support, physical activity and adaptive coping), indicating that risk is not uniformly distributed across campuses or degree programmes and that several determinants are actionable within university settings.¹⁵ In parallel, a systematic review of Italian university counselling documents sharp surges in help-seeking for anxiety/depression, rapid adoption of telecounselling and uneven regional coverage, suggesting that outcomes may be mediated as much by service capacity and access as by individual symptom severity.¹⁵ Discipline-specific Italian data deepen this picture: among Italian undergraduates, sleep quality and workload emerge as consistent correlates and predictors of anxiety/depression, highlighting sleep as a low-cost, high-yield target for screening and early intervention embedded in coursework and clinical rotations.¹⁶ Region-focused analyses (eg, Sicily) further link depressive symptom severity tightly to the frequency of suicidal thoughts in medical trainees, supporting scalable symptom-based screening (eg, brief self-reports plus targeted follow-up) to identify high-risk subgroups early and locally.¹⁷ Read alongside multicentre findings in Italian medical students¹⁰ and longitudinal cohorts bracketing the pandemic period,^{11 12} these newer data shift the emphasis from static prevalence to implementation questions: how to prioritise stepped-care pathways where counselling capacity is thinnest, how to integrate sleep-management and stress-management modules in high-pressure curricula and how to deploy routine, brief risk checks (eg, validated ideation items with rapid triage) during predictable stress peaks (exam blocks and clinical rotations). In short, Italy’s evidence base now highlights the need for stratified prevention approaches that combine university-wide well-being promotion with targeted strategies for specific subgroups to alter negative trajectories in the vulnerable minority that would otherwise remain symptomatic after stressors have subsided.^{7 14–16} A recent systematic review synthesised evidence from more than one hundred studies on Italian university students, confirming high prevalence rates of anxiety and depression but also highlighting substantial methodological heterogeneity and a lack of longitudinal designs, underscoring the need for prospective multicentre studies such as the Survey on Anxiety, Depression, Internalising and Externalising symptoms (SADIE).¹⁸

Maladaptive perfectionism

Recent studies indicate that maladaptive perfectionism has been steadily increasing among young adults and is particularly salient in academic contexts where high standards and external evaluations are pervasive.¹⁹ The Perfectionism Social Disconnection Model asserts that perfectionism erodes self-esteem, fosters social alienation and heightens risk for internalising psychopathology and suicidal ideation.^{20 21} Empirical evidence in university students indicates that perfectionistic concerns are associated with higher levels of anxiety and depression and an increased likelihood of suicidal thoughts, even when controlling for baseline depressive and anxiety symptoms.²² Italian data echo this pattern in high-achieving cohorts: when intense academic stress and repetitive negative thinking coexist with elevated perfectionistic concerns, trajectories of internalising symptoms tend to worsen over time, supporting the inclusion of perfectionism constructs among modifiable targets in prevention frameworks.²³

Maladaptive coping behaviours

In parallel, maladaptive coping behaviours, notably hazardous alcohol and tobacco use, are pervasive among students seeking relief from academic stress.²⁴ Furthermore, students with high alcohol consumption (elevated Alcohol Use Disorders Identification Test-3 (AUDIT-3) scores) or nicotine dependence (high Cigarette Dependence Scale-5 (CDS-5) scores) tend to display greater academic burnout and more persistent depressive symptoms over time compared with low-risk peers.²² Beyond substance use, certain lifestyle vulnerabilities, particularly poor sleep quality, are strongly associated with affective symptoms in university students. A recent Italian study among nursing students confirmed that sleep disturbances and other health behaviours are significantly correlated with higher levels of anxiety and depression,¹⁶ suggesting that interventions targeting these factors (eg, improving sleep) may benefit students’ mental health.

Suicidal ideation and behaviours

Suicidal ideation and behaviours remain a pressing public health concern within university populations. Annual prevalence estimates range from 10% to 15% for suicidal thoughts, with a smaller subset progressing to planning or attempts.⁶ Suicidal risk is particularly elevated among students who exhibit high levels of perfectionism, repetitive negative thinking (eg, rumination and worry) and low social support.^25–27 Cross-lagged and ideation-to-action frameworks highlight that cognitive–affective vulnerabilities may drive the emergence of suicidal ideation, whereas volitional moderators and capability factors influence the progression to behaviour.^{28 29}

Adding geographically informative data, a recent study conducted in Southern Italy (Sicily) observed high levels of depressive and anxiety symptoms and found that Depression, Anxiety and Stress Scales–21, particularly depression, robustly predicted the frequency of suicidal thoughts among medical students, underscoring both the magnitude of the problem and the feasibility of scalable risk screening within Italian faculties.¹⁷ These findings complement national and international evidence linking symptom severity with suicidal ideation and indicate that region-specific monitoring can identify local needs for prevention and support.^{25 30}

Gap in the literature

Despite the wealth of cross-sectional studies, longitudinal research capturing the dynamic interplay between individual vulnerabilities, pandemic-related stressors and mental health outcomes remains scarce in Italy. Recent national syntheses call for prospective, multisite designs that can disentangle transient from persistent symptom courses and test modifiable predictors across multiple academic years.^{14 15} The SADIE protocol addresses this gap through a multicentre, prospective design with four assessment waves over 36 months, integrating validated Italian adaptations of psychometric instruments and mixed-effects and growth-mixture analytical frameworks.

Guiding longitudinal studies

Recent longitudinal research has provided valuable insights into the trajectories of university students’ mental health by examining various contextual, psychological and academic predictors. For instance, studies tracking cohorts across the COVID-19 pandemic demonstrated acute spikes in depressive and anxiety symptoms during restrictions, followed by post-lockdown remission, while highlighting the detrimental role of excessive social media use.^{11 31} Other longitudinal investigations have focused on specific psychological risk and protective factors; higher university connectedness was found to predict a lower incidence and faster remission of internalising symptoms,³² whereas declining self-esteem and elevated academic or financial worries consistently predicted worsening mental health outcomes.^{27 33} Furthermore, the interplay between academic achievement and psychological distress has been shown to be bidirectional, with burnout predicting subsequent declines in grade point average (GPA) and vice versa, a relationship heavily modulated by personality traits such as grit.³⁴

Despite these valuable insights, existing longitudinal research suffers from collective limitations that leave a critical knowledge gap. Most of these investigations are geographically restricted to single-institution cohorts, confined to specific cross-cultural contexts outside of Italy or temporally bound to specific historical events like the pandemic.^{11 31 33 35} Furthermore, they typically isolate narrow psychological or performance variables (such as self-esteem or GPA) rather than comprehensively tracking the concurrent evolution of both internalising and externalising symptoms over an extended timeframe.^32–34 Consequently, there is a lack of large-scale, nationwide, multicentre longitudinal data capable of mapping heterogeneous symptom trajectories using advanced statistical modelling. SADIE directly addresses this gap by implementing a 36-month, four-wave tracking protocol across 14 diverse Italian universities, providing the robust, multi-dimensional framework necessary to overcome these collective limitations.

Study overview and objectives

Objectives

SADIE aims to address the current lack of large-scale longitudinal data on the mental health of Italian university students through a multicentre prospective design. Specifically, the study will:

1. Estimate the prevalence and incidence of anxiety, depression, burnout and suicidal ideation over a 36-month period.

2. Identify psychological, behavioural and contextual predictors of adverse mental health trajectories, including multidimensional perfectionism, hazardous alcohol use, tobacco dependence, academic stress and repetitive negative thinking.

3. Examine whether academic stress and repetitive negative thinking moderate the relationship between perfectionism and subsequent mental health outcomes.

Hypotheses

H1. Higher levels of socially prescribed perfectionism will predict longitudinal increases in anxiety, depressive symptoms, burnout and suicidal ideation.

H2. Hazardous alcohol use and tobacco dependence will be associated with more persistent and severe symptom trajectories over time.

H3. Academic stress and repetitive negative thinking will strengthen the association between perfectionistic concerns and adverse mental health outcomes.

H4. We hypothesise that growth-mixture modelling (GMM) will identify at least three distinct trajectory classes (resilient, transient and chronic).

Methods

Study design

This prospective, multicentre, observational cohort study is designed to track the trajectories of anxiety, depression, burnout and associated risk factors among Italian university students over 36 months. Data collection will occur at four time points: baseline, 6-month, 18-month and 36-month follow-ups. The study spans 14 Italian universities (University of Padova, Aquila, Bologna, Cagliari, Campania, Catania, Catanzaro, Parma, Perugia, Piemonte Orientale, Salerno, Siena, Trieste and Udine), selected to ensure geographic and demographic diversity. Recruitment and data collection at each site will commence only after obtaining approval from their respective local ethics committees. SADIE has already received formal approval from the Ethics Committee of the Department of General Psychology at the University of Padova (Comitato Etico Area 17, Protocol Number 235-d) on 25 October 2024.

Participants

A total of 5000 students enrolled in undergraduate or graduate programmes at participating universities will be recruited for the study. To be eligible, participants must be at least 18 years old, currently enrolled in an undergraduate or graduate programme and fluent in either Italian or English, since the questionnaires are available in both languages to facilitate the recruitment of foreign students. To reduce attrition bias and maintain temporal consistency, individuals who do not complete the baseline questionnaires or who submit follow-up surveys >30 days after the reminder will be excluded. Additionally, students who have already completed an undergraduate or graduate degree will not be eligible. The target sample size assumes an initial invitation to ~10 000 students with a 50% response rate (yielding 5000 participants) and an anticipated 30% attrition rate at each follow-up, based on Meda et al.¹¹ A priori power analysis, informed by effect sizes in similar cohorts, was conducted using R (v.4.3.1)³⁶ and the simr package for Monte Carlo power simulation. The model assumed an intra-subject intraclass correlation coefficient (ICC) of approximately 0.75 for Beck Depression Inventory–II (BDI-II) total scores across assessment waves, consistent with test–retest reliability reported in comparable longitudinal student samples. The simulation specified random intercepts for participants and random slopes for time (slope variance=0.05), with α=0.05 (two-sided). Under these assumptions and accounting for 30% attrition at each follow-up wave, the analysis indicated that this design provides ≥80% power to detect small-to-medium effect sizes (Cohen’s d=0.20–0.50) in our mixed-effects trajectory models.

Instruments

Sociodemographic schedule

Participants will complete a concise sociodemographic questionnaire covering key background variables (gender, age, sexual orientation, language proficiency and ethnicity), academic and occupational characteristics (primary occupation, training year for medical trainees, institution of affiliation, highest educational qualification, faculty area and secondary employment status with motivations) and work-related and lifestyle factors. Work-related measures include perceived work overload (frequency of feeling overwhelmed), number of monthly night shifts and post-shift functioning, weekend work frequency, average weekly working hours, tendencies for early arrival and late departure, collegial cooperation versus competition and weekly study hours outside of work. Conditional logic (eg, study year shown only to trainees) is handled by REDCap branching to streamline the participant experience.

Participants will also provide detailed clinical history information. This includes lifetime and current mental health diagnoses (eg, mood, anxiety, psychotic and personality disorders), family history of psychiatric disorders and degree of kinship, history of self-harm without suicidal intent (lifetime and past 6 months), history and current use of psychotherapy (including frequency), history and current pharmacotherapy for mental health (with type of medication) and history of physical health conditions requiring ongoing care or medications. Physical health measures cover the number and type of diagnosed medical conditions and related treatments. These items use a mix of checkbox, radio and text-entry fields with appropriate branching logic to capture detailed clinical profiles while minimising respondent burden.

A summary of outcomes, endpoints and instruments is presented in table 1.

Table 1. Outcomes, endpoints and instruments.

Outcome	Endpoint	Instrument
Anxiety symptoms	Total score on GAD-7	Generalised Anxiety Disorder Scale (GAD-7)
Depressive symptoms	Total score on BDI-II	Beck Depression Inventory-II (BDI-II)
Emotional exhaustion	Average score on MBI-2	Maslach Burnout Inventory-2 (MBI-2)
Well-being	Sum score on WHO-5	WHO-5 Well-being Index
Hazardous alcohol use	Score ≥5 on AUDIT-3	Alcohol Use Disorders Identification Test-3 (AUDIT-3)
Tobacco dependence	Score ≥3 on CDS-5	Cigarette Dependence Scale-5 (CDS-5)
Suicidal ideation	Positive responses on C-SSRS	Columbia Suicide Severity Rating Scale (C-SSRS)
Perfectionism	Subscale scores on HF-MPS	HF Multidimensional Perfectionism Scale

Maslach Burnout Inventory-2 (original version by Maslach and Jackson; two-item validation by Li-Sauerwine et al, Italian validation by Sirigatti et al)

A brief two-item version of the Maslach Burnout Inventory (MBI), measuring emotional exhaustion and depersonalisation on a 7-point Likert scale (0=never to 6=every day). MBI-2 has demonstrated a correlation (Spearman’s ρ=0.65) with the full MBI and has been used as a rapid screening tool. Items assess the frequency of feeling emotionally drained and cynical towards academic work. The scale has been translated for this study; no formal Italian validation is available.^37–39

Beck Depression Inventory-II (original version by Beck et al, Italian validation by Ghisi et al)

A 21-item self-report inventory evaluating depressive symptoms over the past 2 weeks. Each item is rated 0–3, with higher total scores indicating more severe depression. Italian university samples report excellent internal consistency (α=0.91) and convergent validity with clinical diagnoses.^{40 41}

Generalised Anxiety Disorder-7 (original version by Spitzer et al, Italian validation by Bolgeo et al)

A seven-item self-report scale assessing generalised anxiety symptoms over the previous 2 weeks, rated on a 4-point Likert scale from 0 (‘not at all’) to 3 (‘nearly every day’). Italian validation in a coronary heart disease sample supports its unidimensional structure, good construct validity, high internal consistency (Cronbach’s α ≈ 0.89) and satisfactory measurement invariance across sex and age groups^{42 43}

WHO-5 Well-being Index (original version by WHO, Italian validation by Carrozzino et al)

A five-item scale assessing positive mood, vitality and general interest in activities during the past 2 weeks. Each item is scored 0–5, with higher scores reflecting better well-being. WHO-5 is widely adopted in primary care and student mental health research, with an Italian test–retest reliability of r=0.76.^{44 45}

Alcohol Use Disorders Identification Test-3 (original version by Bush et al, Italian validation by Struzzo et al)

A shortened three-item version of the Alcohol Use Disorders Identification (AUDIT) screening for hazardous alcohol consumption. Items cover frequency of drinking, typical quantity and frequency of heavy episodic drinking. A threshold score ≥5 indicates risky use. AUDIT-3 maintains robust concordance (κ=0.78) with diagnostic interviews and is recommended in digital survey formats.^{46 47}

Cigarette Dependence Scale-5 (original version by Etter et al, Italian adaptation for this study)

A five-item measure of tobacco dependence assessing craving, withdrawal and smoking patterns on a 1–5 scale. CDS-5 demonstrates high internal consistency (α=0.85) and has been adapted into Italian for this study. The scale has been translated for this study; no formal Italian validation is available.⁴⁸

Columbia-Suicide Severity Rating Scale (original version by Posner et al, Official Italian translation used for this study)

A six-item tool evaluating the severity and intensity of suicidal ideation and behaviours. Each question elicits dichotomous (yes/no) responses, supplemented by follow-up severity ratings. Columbia-Suicide Severity Rating Scale (C-SSRS) is highly sensitive and has been culturally adapted to Italian contexts, with automated REDCap logic for risk alerts.^{49 50}

Hewitt and Flett Multidimensional Perfectionism Scale (15 items; original version by Hewitt and Flett, Italian validation by Lombardo et al)

Assesses self-oriented, other-oriented and socially prescribed perfectionism across 15 items, rated on a 7-point Likert scale. Italian cohorts report excellent reliability (α=0.88–0.91). The Hewitt and Flett Multidimensional Perfectionism Scale (HF-MPS) subscales provide nuanced insights into perfectionistic dimensions and are formatted with clear item grouping.^{21 51}

Primary outcomes, predictors and covariates

In this protocol, the primary outcomes are continuous symptom and well-being measures: depressive symptoms (BDI-II total score), anxiety symptoms (GAD-7 total score), emotional exhaustion/burnout (MBI-2 mean score) and psychological well-being (WHO-5 total score). In this protocol, suicidal ideation (C-SSRS) is treated as a co-primary outcome, in line with our hypotheses and with its role as a key endpoint of the study, and is therefore listed alongside BDI-II, GAD-7, MBI-2 and WHO-5. Substance-use measures (AUDIT-3 and CDS-5) have a dual analytic role, serving both as time-varying predictors of subsequent affective and burnout outcomes and as secondary outcomes. Secondary outcomes include hazardous alcohol use (AUDIT-3), tobacco dependence (CDS-5), suicidal ideation and behaviours (C-SSRS) and multidimensional perfectionism (HF-MPS subscale scores), analysed either as continuous or categorical variables depending on the specific research question.

The main psychological and behavioural predictors of interest are multidimensional perfectionism (HF-MPS), academic stress (self-reported workload and overload indicators), repetitive negative thinking (worry and rumination items), hazardous alcohol use (AUDIT-3), tobacco dependence (CDS-5), sleep quality and other lifestyle factors. Sociodemographic and academic characteristics (eg, age, gender, sexual orientation, university, degree programme and training stage), clinical history (prior mental health diagnoses and treatment) and physical health conditions will be treated as potential confounders in multivariable models. In line with our hypotheses, academic stress, repetitive negative thinking and academic group (undergraduate, postgraduate, PhD and medical residents) will be examined as pre-specified effect modifiers of the association between perfectionism and mental health trajectories

Procedure

The 13 participating universities other than Padova were recruited by contacting student representatives and, through them, assessing the interest of psychiatry or clinical psychology professors, who were subsequently designated as local referents. Data collection follows a staggered timeline across these institutions, spearheaded by the University of Padova as the coordinating centre, which obtained the primary ethical approval and initiated baseline recruitment in October 2024. Among the currently active sites, data collection commenced at the University of Udine in April 2025, followed by the University of Campania in March 2026 and the University of Salerno, University of Trieste and University of Eastern Piedmont in May 2026. The baseline recruitment and data collection period is currently ongoing, as the remaining participating universities are progressively completing their local ethics committee approval processes for sequential activation.

Recruitment across all active centres will employ a convenience sampling, open-invitation approach rather than a random sampling strategy. Dissemination will involve institutional emails containing open REDCap survey links (provided in separate Italian and English versions) as well as on-site and classroom presentations at the participating universities, allowing any interested student to participate on a voluntary basis. After providing informed consent online, participants will complete the baseline assessment (≈30 min). To verify student enrolment status and ensure precise institutional tracking, the initial socio-demographic section of the questionnaire includes mandatory self-report fields where respondents must explicitly specify their affiliated university and their specific degree programme. Entries that do not meet these academic criteria will be screened out during the data cleaning phase.

Surveys will be administered longitudinally across four distinct waves: at 0 (baseline), 6, 18 and 36 months. For each individual participant, the total period of data collection and longitudinal follow-up spans exactly 36 months from their initial entry. To facilitate this tracking, the REDCap platform is configured to use the email addresses voluntarily provided by participants during the baseline survey to automatically schedule and dispatch timed follow-up reminders based strictly on the individual’s baseline completion date. To ensure confidentiality, participant email addresses will be pseudo-anonymised as unique alphanumeric codes (eg, RECORD_ID_N), and all data will be stored on General Data Protection Regulation (GDPR)-compliant servers.

As illustrated in figure 1, starting from an expected 5000 baseline participants, we anticipate approximately 3500, 2450 and 1715 completers at 6-month, 18-month and 36-month follow-ups, respectively, corresponding to an attrition rate of around 30% at each wave, consistent with previous longitudinal data in similar cohorts.¹¹

Figure 1. Flowchart of participant recruitment, inclusion, follow-up and analysis. The diagram shows the expected proportions of participants at each stage: initial pool, eligibility assessment, enrolment, completion of baseline questionnaires and retention across the three planned follow-ups (6, 18 and 36 months). Estimated attrition rates are reported at each step.

Statistical analysis

Primary outcomes include the prevalence and incidence of mental health symptoms, calculated with 95% CIs.

For primary analyses, symptom and well-being measures (BDI-II, GAD-7, MBI-2 and WHO-5) will be treated as continuous outcomes. In addition, for descriptive purposes and specific secondary analyses, we will derive binary indicators of ‘probable disorder’ using cut-off scores that are widely adopted in the original and validation literature of these instruments. Specifically, depressive symptoms will be categorised according to standard BDI-II severity thresholds (eg, scores in the moderate or higher range) and anxiety according to established GAD-7 severity bands (eg, scores ≥10 indicating at least moderate anxiety). Reduced psychological well-being will be defined using the recommended WHO-5 threshold indicating poor well-being and the need for further clinical assessment. Hazardous alcohol use will be operationalised as an AUDIT-3 score at or above the commonly used risk cut-off and at least moderate tobacco dependence as a CDS-5 score at or above the threshold adopted in previous studies of cigarette dependence. Suicidal ideation and behaviours will be coded according to standard C-SSRS categories, distinguishing any suicidal ideation from high-risk ideation (with intent and/or plan) and suicidal behaviour.

Secondary analyses will employ generalised linear mixed-effects models (GLMM) to evaluate symptom trajectories over time, adjusting for covariates such as perfectionism and substance use. Logistic regression will identify predictors of adverse outcomes (eg, dropout rates and tobacco use). All analyses will be conducted using R (V.4.3.1)³⁶ with packages such as lme4 for mixed models and ggplot2 for data visualisation. To mitigate potential self-selection and attrition bias, several complementary strategies will be adopted. First, post-stratification weights will be computed so that the joint distribution of key socio-demographic and academic variables in our analytic sample approximates national enrolment patterns in Italian universities (eg, age, gender, institution and degree level). Second, for each follow-up wave, the probability of participation will be estimated as a function of baseline socio-demographic, academic and clinical characteristics, and inverse probability weights will be derived for follow-up that will be incorporated into longitudinal models to account for differential loss to follow-up. Third, all mixed-effects models will be fitted using full information maximum likelihood estimation to make use of partially observed data under a missing-at-random assumption. As sensitivity analyses, we will compare weighted with unweighted estimates, explore alternative specifications of the missingness model and examine worst-case scenarios for attrition to assess the robustness of prevalence and trajectory estimates.

Ethical and dissemination

This study will be conducted in accordance with the ethical principles outlined in the Declaration of Helsinki and has already received formal approval from the Ethics Committee of the Department of General Psychology at the University of Padova (Comitato Etico Area 17, Protocol Number 235-d). Before initiating data collection at each of the 14 participating universities, local ethics committees will review and approve the protocol to ensure compliance with institutional and national regulations.

Participants will be thoroughly informed about the study’s objectives, procedures and their rights through a digital consent process integrated into the REDCap platform. Consent will be obtained electronically, emphasising the voluntary nature of participation and the freedom to withdraw at any time without penalty. To safeguard participant well-being, robust risk management protocols will be implemented. For instance, participants reporting suicide ideation with (probable) intentionality (BDI-II item 9=3) or plan/preparatory behaviour (C-SSRS item 5 or C-SSRS item 6 positive)⁴⁹ will receive automated alerts with direct links to local mental health services.

Data security will be prioritised at every stage. Participant identities will be protected through pseudo-anonymisation, with email addresses replaced by unique alphanumeric codes (RECORD_ID_XXXX). All data will be stored on GDPR-compliant servers managed by the University of Padova, accessible exclusively to authorised team members.

Transparency and accountability are central to the study’s design. Investigators have disclosed no financial or non-financial conflicts of interest, and the protocol adheres rigorously to the STROBE guidelines for observational cohort studies. By integrating these ethical safeguards, the study aims to strike a balance between rigorous scientific inquiry and unwavering respect for participant dignity and safety.

Study findings will be disseminated through peer-reviewed publications and presentations at national and international conferences. In addition, aggregated summary reports will be shared with participating universities and relevant student services to inform local prevention strategies. Subject to institutional and data protection regulations, deidentified datasets and analysis scripts may be made available on reasonable request to support transparency and secondary analyses.

Discussion

SADIE could represent a step forward in the understanding of student mental health. While current evidence consistently documents high prevalence rates of anxiety, depression, burnout and suicidal ideation among university students, most available research remains static, fragmented and cross-sectional. This methodological limitation prevents us from distinguishing transient distress related to predictable academic stressors (such as exams or transitions between training stages) from persistent, clinically relevant symptom trajectories that may lead to chronic impairment or suicidal outcomes. SADIE directly addresses these limitations by following 5000 students across 14 universities for 36 months, employing repeated assessments, validated psychometric tools and advanced statistical models. This design will provide a rare opportunity to map the dynamic trajectories of student well-being over time, delivering both national and international insights.

The multicentre nature of SADIE ensures diversity across geographical regions, academic fields and institutional settings, capturing the heterogeneity of the Italian higher education system and increasing the generalisability of findings. By recruiting from medical, health and non-health faculties alike, SADIE will also enable discipline-specific analyses, clarifying whether the well-documented vulnerabilities of medical and healthcare students extend to or differ from those observed in other fields. Such granularity has been largely missing in Italian literature to date.

The study’s theoretical contribution lies in its systematic focus on maladaptive perfectionism, repetitive negative thinking and academic stress as core risk factors, measured through the HF-MPS and targeted items on worry and rumination. Alongside these constructs, SADIE incorporates established measures of affective symptoms—BDI-II for depression, GAD-7 for anxiety and WHO-5 for positive psychological functioning—allowing us to model the interplay between vulnerability factors and mental health outcomes with precision. While the Perfectionism Social Disconnection Model has long posited that perfectionism erodes self-esteem and fosters alienation, SADIE will explicitly test this hypothesis in relation to both negative outcomes (BDI-II, GAD-7 and suicidal ideation via C-SSRS) and protective dimensions (WHO-5), providing a balanced perspective on risk and resilience.

In addition, SADIE integrates behavioural and lifestyle risk factors that are particularly salient in student populations, including hazardous alcohol use (AUDIT-3), tobacco dependence (CDS-5), sleep disturbance and low levels of physical activity. These behaviours often represent maladaptive coping strategies for academic stress and may interact with underlying affective vulnerability. At the same time, burnout-assessed through the MBI-2 will serve as both an outcome and a mediator, linking stress exposure and coping behaviours to worsening of mental health. The longitudinal interaction between these dimensions and symptom measures (BDI-II, GAD-7 and WHO-5) has rarely been investigated in large student cohorts, and SADIE will generate comprehensive evidence to clarify these pathways.

From a methodological perspective, SADIE brings together several strengths that position it uniquely in the international landscape:

• A large, nationally representative sample, ensuring statistical power to detect small-to-medium effect sizes and explore subgroup differences.

• A multicentre design across 14 universities, allowing for institutional comparisons and highlighting potential structural determinants of mental health.

• Use of validated and internationally comparable psychometric measures, alongside carefully adapted Italian versions where necessary.

• Four timepoints over three academic years, enabling the study of both short-term fluctuations and long-term persistence.

• Application of advanced modelling approaches, including GLMM and GMMs, to capture heterogeneity in symptom trajectories and identify high-risk latent subgroups.

The expected contributions of SADIE extend beyond descriptive epidemiology. By clarifying how perfectionism, stress, coping behaviours and social contexts interact over time to influence symptom persistence, the study will advance mechanistic understanding of student mental health. This knowledge has immediate translational value: universities will be able to identify subgroups of students most at risk, allocate limited counselling resources more effectively and implement stratified prevention strategies tailored to specific vulnerabilities. In particular, the integration of suicidal ideation as a key endpoint will support the development of scalable, evidence-based screening and triage protocols—an urgent need given the documented rise in suicide risk among young adults.

At the national level, SADIE will generate data that can directly inform university counselling services, student well-being initiatives and mental health policy. Findings will provide evidence for the integration of prevention programmes within curricula, the adoption of routine mental health checks during predictable stress peaks and the prioritisation of sleep and stress management modules in high-pressure academic tracks such as medicine. At the international level, SADIE will contribute as one comprehensive longitudinal dataset on European students, offering a benchmark for cross-cultural comparisons and a template for future multicentre studies.

In summary, SADIE is expected to make three major contributions to the literature and practice:

1. Filling a critical methodological gap by moving from cross-sectional prevalence studies to prospective, dynamic models of student mental health.

2. Advancing theoretical understanding by testing perfectionism-based and stress-based models of vulnerability in a large, diverse and longitudinal cohort.

3. Delivering actionable evidence for prevention, policy and service delivery in higher education, both in Italy and internationally.

By combining validated instruments—BDI-II, GAD-7, WHO-5, MBI-2, C-SSRS, HF-MPS, AUDIT-3 and CDS-5—within a prospective multicentre design, SADIE will not only quantify prevalence and incidence but also test mechanistic models of vulnerability and resilience. This integration ensures comparability with international research, enhances the robustness of findings, and provides direct translational value for prevention frameworks.

Limitations and dissemination

Anticipated limitations

While SADIE benefits from a robust multicentre design, several methodological limitations must be anticipated, particularly regarding their potential impact on the study’s findings. First, the voluntary nature of our recruitment strategy and the use of convenience sampling via open REDCap links introduce a risk of self-selection bias. This could impact our findings in two contrasting ways: students experiencing higher levels of psychological distress might be more motivated to participate to seek validation or visibility, leading to an overestimation of baseline prevalence rates; conversely, those with severe functional impairment due to mental health issues might avoid the survey entirely, causing an underestimation. Consequently, caution will be required when generalising the absolute baseline prevalence data to the entire Italian university population.

Second, tracking a cohort over a 36-month timeframe across four distinct waves introduces a substantial risk of participant attrition and loss to follow-up. In longitudinal mental health surveys, dropout is rarely missing completely at random. There is an inherent risk of a ‘healthy survivor’ effect, in which students who are experiencing worsening internalising or externalising symptoms and academic burnout or those who drop out of university entirely become progressively less likely to complete follow-up waves. This potential selective attrition could lead to an artificial flattening or underestimation of worsening symptom trajectories over time.

Third, as SADIE is a prospective observational cohort rather than an experimental study, the design does not allow for strict causal inference regarding the relationships between predictors and mental health outcomes. Associations and trajectories will therefore need to be interpreted in terms of risk patterns and correlational evidence rather than causality.

Fourth, although all instruments were selected based on their strong international validation and, where available, Italian psychometric data, two measures used in SADIE (MBI-2 and CDS-5) do not yet have a formal Italian validation. This limits the ability to compare absolute cut-offs and dependence thresholds with other Italian samples and requires cautious interpretation of these specific scales.

To mathematically counteract these potential impacts and safeguard the validity of our findings, our statistical framework incorporates specific mitigation strategies in alignment with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. We will implement post-stratification weights to realign our sample’s socio-demographic and academic distribution with broader national university metrics and employ inverse probability weighting (IPW) alongside full information maximum likelihood estimation within our GMMs to structurally account for non-random missingness over time.

Dissemination plan

• Results will be published in open-access journals.

• Workshops will be organised at participating universities to translate findings into campus mental health policies.

Roles and responsibilities

The coordinating centre (University of Padova, Department of General Psychology) is responsible for overall study design, data management and analysis. Local investigators at each participating site will oversee recruitment, ethics approval and participant support. Data collection will be centralised via REDCap. All investigators will contribute to interpretation of results and dissemination.