Vietnamese Diabetes Study Puts a Famous Anxiety and Depression Test Under the Microscope

Type 2 diabetes affects close to 600 million people worldwide, and the burden it places on the mind is nearly as heavy as the one it places on the body. Anxiety and depression are strikingly common among people living with the condition, and both make it harder to take medications on schedule, eat well, and keep blood sugar under control. Yet detecting that emotional burden requires a measurement tool that actually works in the language and culture of the patient being assessed. A new study from Vietnam, published in BMC Psychology, has now put one of the world’s most widely used psychological questionnaires through a rigorous statistical stress test in Vietnamese adults with type 2 diabetes, and the results are both reassuring and cautionary.

The instrument in question is the Hospital Anxiety and Depression Scale, or HADS, a fourteen-item questionnaire first developed in the 1980s by R.P. Snaith and A.S. Zigmond. Its genius lies in what it leaves out. Most depression questionnaires ask about fatigue, appetite loss, or sleep problems, but in a hospital ward or a diabetes clinic those symptoms may reflect the underlying illness rather than a mood disorder. HADS was deliberately designed to avoid physical symptoms, focusing instead on the emotional core of anxiety and depression, which is why it has become a staple in chronic disease settings around the globe. Until now, however, nobody had formally evaluated how well the Vietnamese version of the scale performs in adults with type 2 diabetes, leaving clinicians and researchers in Vietnam without clear guidance on how to interpret its scores.

To close that gap, a team led by Son Huynh of the University of Medicine and Pharmacy at Ho Chi Minh City recruited 465 adults with type 2 diabetes from three different outpatient diabetes clinics in Ho Chi Minh City. The multi-site design was not a luxury; it was a deliberate choice to see whether the questionnaire behaves consistently across different clinical settings, something single-site validation studies cannot reveal. Each participant completed the Vietnamese HADS along with the Patient Health Questionnaire-9, a well-established depression measure that served as a benchmark for comparison. The study was approved by the Ethics Committee for Biomedical Research at the University of Medicine and Pharmacy at Ho Chi Minh City, and all participants provided written informed consent.

The first question the researchers tackled was reliability: does the scale produce consistent, stable measurements? Using modern ordinal reliability coefficients, which are appropriate for the ordered response categories typical of psychological questionnaires, they found an ordinal McDonald’s omega of 0.765 for the anxiety subscale and 0.846 for the depression subscale. In practical terms, both figures fall within the range generally considered acceptable to good, meaning patients’ answers hang together coherently enough to support meaningful score interpretation. That is a foundational result, because no amount of sophisticated structural analysis can rescue a questionnaire whose items fail to measure anything consistently.

The deeper question was structural validity: does the questionnaire actually separate into the two distinct dimensions, anxiety and depression, that its design promises? The answer turned out to be more complicated than a simple yes. Exploratory factor analysis, guided by parallel analysis, initially suggested three underlying factors rather than two. But when the team inspected the third factor, they found it was fragmenting the depression items without representing a genuinely distinct psychological construct, so they retained the original two-factor solution. Confirmatory factor analysis then confirmed that two-factor structure, though only after the researchers incorporated two theory-driven correlations between residual terms, adjustments that account for overlapping wording in the Vietnamese translation. The final model showed acceptable fit, with a root mean square error of approximation of 0.084, a comparative fit index of 0.922, a Tucker-Lewis index of 0.904, and a standardized root mean square residual of 0.076.

The most technically sophisticated part of the analysis was a bifactor model, a statistical framework that asks whether a questionnaire is best understood as measuring one broad underlying dimension plus some specific subdimensions, or as several largely independent dimensions. The results were striking. A substantial general distress factor emerged, explaining a large share of the common variance: the omega hierarchical coefficient reached 0.839 and the explained common variance was 0.608. In plain language, most of what the Vietnamese HADS measures in this population is a single, broad current of psychological distress that runs through both the anxiety and depression items. At the same time, enough residual variance remained within each subscale to suggest that the anxiety and depression scores still carry some distinct, meaningful information beyond the general factor.

Construct validity testing added further nuance. Both subscales correlated moderately with the PHQ-9, supporting convergent validity, the idea that the scale is measuring something related to what similar tools measure. Discriminant validity, the ability of the two subscales to keep themselves statistically distinct, was supported for the depression subscale but not for the anxiety subscale. The square root of the average variance extracted for anxiety was 0.552, which fell below the correlation between the two subscales of 0.652. That means the anxiety and depression scores overlap so heavily in this population that they cannot be treated as cleanly separable constructs, echoing the bifactor finding that general distress dominates the measurement.

Perhaps the most practically important discovery came from an exploratory analysis of cross-site comparability. When the researchers compared item responses across the three clinics, three depression items related to anhedonia, the loss of interest or pleasure in activities, showed comparatively larger differences between sites, with eta-squared values ranging from 0.18 to 0.21. Those are sizable effects, indicating that responses to these particular questions varied noticeably depending on which clinic a patient attended. The authors caution that this heterogeneity means researchers should be careful when comparing or pooling HADS scores across different settings in multi-site studies, because some of the observed differences between sites may reflect measurement inconsistency rather than genuine differences in patient wellbeing.

So what should clinicians and researchers in Vietnam actually do with the Vietnamese HADS? The study’s conclusions are deliberately measured. The scale demonstrates acceptable structural validity and partial construct validity in Vietnamese adults with type 2 diabetes, and the bifactor results support the total score as a broad, cross-sectional index of general psychological distress. The authors recommend that the total score should be the primary measure reported, with the individual subscale scores offering useful but more limited supplementary information. Crucially, diagnostic accuracy and criterion validity, the ability of the scale to identify true clinical cases against a diagnostic gold standard, have not yet been established. The total score is therefore supported strictly as a continuous indicator of emotional burden for research and routine care, not as a formal screening or diagnostic tool.

The study also carries a broader lesson for global health measurement. Patient-reported outcome measures do not travel effortlessly across languages and cultures; every translation inherits quirks of wording, and every new clinical population can reshape how items behave. By combining classical reliability estimation, exploratory and confirmatory factor analysis, bifactor modelling, and multi-site heterogeneity checks, the Vietnamese team has modeled the kind of thorough validation that questionnaires deserve before they are used to inform care decisions. For the hundreds of millions of people worldwide living with type 2 diabetes, and for the clinicians trying to see beyond the disease to the person, having a distress measure that is trusted, transparent about its limits, and grounded in local evidence is a quiet but consequential step forward for mental health care in chronic disease.

Subject of Research: Psychometric validation of the Vietnamese Hospital Anxiety and Depression Scale in adults with type 2 diabetes

Article Title: Psychometric validation of the Vietnamese Hospital Anxiety and Depression Scale (HADS) in adults with type 2 diabetes mellitus: a multi-site cross-sectional study

Article References: Huynh, S., Pham, O., Nguyen, N., Phan, L., Le, T., Nguyen, V., & Pham, A. (2026). Psychometric validation of the Vietnamese Hospital Anxiety and Depression Scale (HADS) in adults with type 2 diabetes mellitus: a multi-site cross-sectional study. BMC Psychology. https://doi.org/10.1186/s40359-026-05608-w

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05608-w

Keywords: HADS, type 2 diabetes, psychometric validation, anxiety, depression, Vietnam, bifactor analysis, psychological distress, patient-reported outcome measures, factor analysis, mental health screening, Ho Chi Minh City