# Psychometric Behaviour of the GAD-7 in Medical Students: Structural Stability, Measurement Equivalence and Contextual Sensitivity

## TL;DR

This study examines how well the GAD-7 anxiety scale works for medical students, finding it reliable but with some limitations in measuring anxiety in this specific group.

## Contribution

The study evaluates the GAD-7's psychometric properties in medical students, focusing on structural stability and measurement invariance in a culturally adapted version.

## Key findings

- The unidimensional structure of the GAD-7 was statistically coherent in medical students.
- Reliability coefficients were high, and convergent validity was supported.
- Measurement invariance was confirmed at the configurational and scalar levels but not fully at the metric level.

## Abstract

Background: Anxiety symptoms among medical students often emerge at the intersection of sustained academic pressure, anticipatory uncertainty and early professional socialisation, complicating their distinction from transient stress responses. Instruments employed in this context are therefore expected to operate consistently across subgroups while preserving conceptual clarity under non-clinical conditions. The Generalized Anxiety Disorder scale (GAD-7), widely adopted as a brief screening measure, has shown variable factorial behaviour across populations, particularly when applied to student cohorts. Materials and methods: Using confirmatory factor analysis with robust weighted least squares estimation, the latent structure of a culturally adapted Spanish version of the GAD-7 was examined in a sample of medical students enrolled across all academic years at a public university. Model performance was evaluated through multiple fit indices suited for ordinal data, alongside estimates of convergent validity based on average variance extracted and reliability assessed via both Cronbach’s α and McDonald’s ω. Measurement invariance across sex was explored through a sequence of increasingly constrained multi-group models. Results: The unidimensional configuration originally proposed for the scale remained statistically coherent, despite minor tensions between absolute and incremental fit indicators commonly reported in comparable university-based samples. Convergent validity estimates suggested that the latent construct accounted for a substantial proportion of item variance, while reliability coefficients fell within the upper range observed internationally. Invariance testing supported comparability at the configurational and scalar levels, although full metric equivalence was less stable. Conclusions: Rather than resolving ongoing debates regarding the internal structure of the GAD-7, these findings situate its psychometric behaviour within the specific demands of medical education, where anxiety-related symptoms may fluctuate between normative adaptation and clinically relevant distress. This positioning invites further examination of how screening instruments perform when anxiety is shaped as much by institutional context as by individual psychopathology.