Digital PROMs, Psychometric Quality and Unsupervised Profiling in a Digital Cohort of University Students: Results From the DiCoBeNe Study

Abstract

Introduction
Digital cohorts represent a relevant infrastructure for the remote and scalable collection of patient-reported outcome measures (PROMs). In university populations, these tools allow for a multidimensional assessment of psychophysical well-being, including sleep quality, perceived stress, anxiety and depressive symptoms, and quality of life. Beyond the evaluation of psychometric properties, the characterization of the latent structure of outcomes through unsupervised approaches is also of particular relevance.

Objectives
To describe the baseline results of the DiCoBENE study (DIgital COhort per la Valutazione del BENEssere degli Studenti Universitari), a digital cohort of university students, pursuing three objectives: to assess data completeness and internal reliability of digital PROMs; to quantify the frequency of adverse outcomes and the associations among the measured domains; and to explore the multidimensional structure of outcomes using latent profile analysis (LPA), principal component analysis (PCA), and network analysis.

Methods
DiCoBENE is an observational, prospective, repeated-measures digital cohort launched in Catania. At baseline, the Pittsburgh Sleep Quality Index (PSQI), Perceived Stress Scale-10 (PSS-10), Generalized Anxiety Disorder-7 (GAD-7), Patient Health Questionnaire-9 (PHQ-9), and WHOQOL-BREF were administered digitally. Analyses included descriptive statistics, Cronbach’s alpha, McDonald’s omega, Spearman correlations, and multivariate analyses using LPA, PCA on standardized scores, and network analysis based on partial correlations.

Results
The baseline sample included 442 participants, with completion rates of 84.2% for GAD-7, PSS-10, and WHOQOL-BREF domains, 83.9% for PHQ-9, and 83.7% for PSQI. The frequency of adverse outcomes was high: poor sleep quality (PSQI > 5) was observed in 58.8%, moderate-to-severe anxiety (GAD-7 ≥ 10) in 46.0%, moderate-to-severe depressive symptoms (PHQ-9 ≥ 10) in 29.5%, moderate stress in 54.0%, and high stress in 19.8%. Internal reliability was good to excellent for GAD-7 (α = 0.877; ω = 0.881), PHQ-9 (α = 0.818; ω = 0.820), and PSS-10 (α = 0.929; ω = 0.931), and moderate for PSQI (α = 0.654; ω = 0.681). Correlations among domains were consistent with theoretical expectations. In particular, PSS-10 correlated with PHQ-9 (rho = 0.759) and GAD-7 (rho = 0.736), while PHQ-9 showed an inverse correlation with the WHOQOL-BREF psychological domain (rho = −0.578). Multivariate analyses, conducted on 365 complete observations, identified an optimal four-profile latent solution through LPA, interpretable as a gradient of psychophysical burden. PCA showed that the first principal component explained 55.6% of the total variance, while the first two components explained 67.4% overall. Network analysis revealed a coherent structure, with greater centrality for WHOQOL-BREF Physical, PHQ-9, WHOQOL-BREF Psychological, and PSS-10.

Conclusions
These findings document the feasibility and methodological robustness of a multidimensional assessment of psychophysical well-being using digital PROMs in a university cohort. The high frequency of adverse outcomes, the adequate reliability of most instruments, and the convergence of findings obtained through LPA, PCA, and network analysis indicate that sample heterogeneity can be described primarily along a continuum of psychophysical burden rather than through distinct phenotypes. These findings support the integrated use of psychometric and unsupervised approaches in the analysis of digital PROMs in complex observational studies.