When Depression, Anxiety and Stress Overlap: What Our DASS-21 Analysis Revealed

Depression, anxiety and stress are usually presented as separate dimensions on the DASS-21, but our analysis of Pakistani university students showed that these domains were very closely connected. The DASS-21 retained an acceptable three-factor structure and good reliability, yet correlations between its underlying depression, anxiety and stress factors were exceptionally high. The finding does not mean the three constructs are identical. It does suggest that, in this student population, psychological distress may be experienced in a highly intertwined way and that subscale scores should be interpreted with appropriate caution.

By Taimoor Asghar

Why the overlap between depression, anxiety and stress matters

University mental health is rarely neat. A student who feels persistently low may also be tense, restless, unable to relax, worried about academic performance and physically agitated. Another student who initially describes the problem as stress may simultaneously experience loss of motivation, hopelessness or symptoms commonly associated with anxiety.

Questionnaires such as the Depression Anxiety Stress Scales-21, or DASS-21, attempt to organize this complexity into measurable domains. The instrument contains three seven-item subscales designed to assess depression, anxiety and stress. In research, the scores are often analyzed separately because each subscale is intended to represent a distinguishable component of emotional distress.

However, establishing three labeled subscales is not the same as demonstrating that those subscales behave as clearly separated psychological constructs in every population. One of the most informative findings from our recent study was therefore not simply whether students had elevated scores. It was how closely the three DASS-21 dimensions were related at the latent level.

Our study, published in BMC Psychology, examined 602 undergraduate students in Lahore, Pakistan, including 424 medical and 178 non-medical students. Alongside comparisons between academic groups, we evaluated the psychometric performance of both the PHQ-9 and DASS-21 using several statistical approaches. Readers can access the full BMC Psychology study on PHQ-9 and DASS-21 psychometric properties in Pakistani university students.

What our DASS-21 analysis examined

The purpose of psychometric analysis is not merely to report whether a questionnaire produces a score. It asks whether the structure and behavior of those scores are consistent with what the instrument is supposed to measure.

For the DASS-21, an important question is whether responses support three related but distinguishable domains: depression, anxiety and stress. We evaluated that question using confirmatory factor analysis, while also examining reliability and differences in scores between medical and non-medical students.

Confirmatory factor analysis, commonly abbreviated CFA, tests how well a proposed measurement structure corresponds to observed response patterns. In practical terms, it allowed us to examine whether the seven depression items tended to reflect a depression factor, the seven anxiety items an anxiety factor and the seven stress items a stress factor in the way expected from the questionnaire’s theoretical structure.

The three-factor model showed acceptable overall fit

The DASS-21 three-factor CFA produced a comparative fit index of 0.954, a Tucker-Lewis index of 0.948 and a root mean square error of approximation of 0.052. Taken together, these indices indicated that the proposed three-domain model represented the observed data reasonably well.

This is an important result because it provides support for retaining the conventional depression, anxiety and stress structure in this sample. The questionnaire was not behaving as though its items were arbitrarily grouped into three unrelated labels.

Reliability was also satisfactory. Across the scales examined in the study, Cronbach’s alpha values ranged from 0.82 to 0.88. These values indicate that the relevant sets of questionnaire items showed good internal consistency in the study population.

Yet the factor analysis revealed another feature that deserves equal attention.

The striking finding: extremely high correlations between DASS-21 domains

Although the three-factor model fitted the data adequately, the latent factors were very strongly correlated. The estimated correlation between depression and stress was 0.939, while the correlation between anxiety and stress was 0.949.

Correlations this high mean that students who were positioned higher on one underlying distress dimension were very likely to be positioned higher on another. In particular, anxiety and stress were extremely closely related in the model.

This creates an important psychometric tension. On one hand, the CFA supported a three-factor organization. On the other, near-unity correlations raise questions about how distinctly those factors were functioning in this particular population.

The correct interpretation is not that the DASS-21 failed, nor that depression, anxiety and stress are interchangeable diagnoses. Rather, the result suggests that the subscales may share a substantial amount of common variance and that distinctions among them may be less pronounced than their separate labels imply.

What does a latent correlation actually mean?

A latent correlation is different from simply correlating two raw questionnaire totals. In a CFA model, each factor represents the common psychological dimension inferred from responses to its corresponding items. The correlation therefore describes the estimated relationship between those underlying factors after the measurement model has been specified.

A high latent correlation tells us that two constructs move together very strongly in the sample. If a student has a high estimated level of anxiety-related distress, for example, the model suggests that the same student is also very likely to have high stress-related distress.

It does not prove that anxiety causes stress, that stress causes depression or that the constructs are clinically identical. Our study was cross-sectional, meaning participants were assessed at one point rather than followed over time. The analysis therefore identifies patterns of association, not temporal or causal pathways.

Why might depression, anxiety and stress overlap so strongly in students?

Several plausible mechanisms could contribute to the overlap, although our data cannot determine which explanation is correct.

1. Emotional distress often crosses diagnostic boundaries

Human psychological experiences do not necessarily follow questionnaire categories. Sleep disruption, cognitive overload, irritability, reduced motivation, difficulty concentrating, tension and persistent negative affect can occur together.

A student facing sustained academic or personal strain may therefore endorse items belonging to several DASS-21 domains during the same period. This shared emotional burden can make statistical separation between subscales more difficult.

2. Depression and anxiety commonly co-occur

Depressive and anxiety symptoms frequently appear together in clinical and non-clinical populations. A person can experience worry, physiological arousal and fear while also reporting low mood, reduced pleasure or hopelessness. Screening instruments capture symptom dimensions rather than enforcing mutually exclusive categories.

For this reason, substantial correlation between questionnaire domains is not inherently surprising. The more notable feature of our study was the magnitude of that overlap, especially for anxiety and stress.

3. Students may experience several pressures simultaneously

Academic deadlines, financial concerns, uncertainty about the future, family expectations, social relationships and sleep disruption can occur concurrently. A single stressful environment may therefore be associated with multiple types of emotional symptoms rather than producing one isolated response.

This is one reason campus mental-health strategies may be more useful when they address broad psychological distress and functioning instead of assuming that students can always be divided cleanly into separate depression, anxiety or stress groups.

4. Measurement itself can contribute to overlap

Psychometric scales simplify complex experiences into structured questions. Even when items are designed for different subscales, the underlying experiences can remain conceptually related. Feeling unable to relax, becoming easily agitated and experiencing intense negative affect may all reflect partially shared distress processes.

High factor correlations can consequently emerge because constructs genuinely overlap, because items capture related aspects of a broader distress dimension, or through a combination of both mechanisms.

Does the overlap mean the DASS-21 should be treated as one scale?

Not automatically.

Our CFA showed adequate fit for the established three-factor model, so the results do not justify simply discarding the depression, anxiety and stress subscales. At the same time, the extremely high latent correlations make it reasonable to question whether every difference between subscale scores represents a sharply distinct psychological construct in this population.

That distinction is important for researchers. A questionnaire may demonstrate good internal consistency and acceptable model fit while still showing limited discriminant separation between closely related factors.

Future studies could examine alternative measurement models, including hierarchical or bifactor approaches in which a broad general distress factor is modeled alongside more specific symptom dimensions. Such analyses could help determine how much of the DASS-21 response pattern reflects common psychological distress and how much reflects uniquely depressive, anxious or stress-related features.

Our study did not establish that a different scoring approach should replace the standard DASS-21 subscales. The appropriate conclusion is narrower: the observed factor correlations suggest that researchers should evaluate and report construct distinguishability rather than relying only on subscale names.

What the DASS-21 results revealed beyond factor correlations

The study also compared symptom scores between medical and non-medical students. Contrary to the assumption that medical students must necessarily carry the greatest psychological burden, non-medical students had higher adjusted scores on the DASS-21 depression and anxiety domains.

In the adjusted regression models, non-medical students scored higher on DASS Depression and DASS Anxiety. The estimated group difference for DASS Stress did not reach conventional statistical significance.

These findings should not be interpreted as evidence that one academic group is universally more vulnerable than another. The study sampled students from universities in Lahore, and academic discipline explained only a small proportion of the overall variation in psychological scores. The regression models had low R-squared values, indicating that most individual differences in symptoms remained unexplained by the included predictors.

That is itself informative. Mental health is shaped by more than the student’s degree program.

Female students reported higher scores across domains

Female gender was associated with higher scores across the psychological domains analyzed in the study. The association remained statistically significant in the adjusted models.

This pattern can help identify population-level differences, but it should not be converted into assumptions about any individual student. Group averages cannot determine whether a particular man or woman is experiencing clinically meaningful distress.

The result instead supports the broader principle that mental-health services should be responsive to differences in burden across student populations while remaining accessible to everyone who needs support.

What high DASS-21 overlap means for screening

Screening tools are most useful when their purpose is clearly understood. The DASS-21 can quantify patterns of self-reported emotional symptoms, but a score is not a clinical diagnosis.

Our findings reinforce several practical principles for researchers, universities and clinicians who encounter DASS-21 results:

• Avoid interpreting subscales in complete isolation. A student with a high anxiety score may also have substantial depressive or stress-related symptoms.
• Consider the broader symptom profile. The three DASS-21 dimensions can provide useful structure, but their strong correlations mean the total pattern may be clinically or academically informative.
• Do not confuse psychometric categories with diagnostic boundaries. The DASS-21 is a symptom questionnaire and should not independently establish a psychiatric diagnosis.
• Evaluate functioning and context. Academic difficulties, sleep, social circumstances, previous mental-health history and other contextual factors may matter alongside questionnaire scores.
• Use elevated scores as a reason for appropriate assessment, not labeling. Screening should facilitate support rather than replace professional evaluation.

Why acceptable reliability is not enough

Mental-health research frequently reports Cronbach’s alpha and then concludes that an instrument is valid because alpha is high. That approach is incomplete.

Reliability concerns consistency. Construct validity concerns whether the instrument measures the theoretical dimensions researchers believe it measures. Factor structure, factor correlations, measurement invariance and other analyses provide information that internal consistency alone cannot supply.

In our sample, DASS-21 reliability was good and the three-factor model fitted adequately. Yet the latent correlations showed that depression, anxiety and stress remained very difficult to separate statistically. Had we reported alpha values alone, that important feature would have been missed.

This illustrates why psychometric studies benefit from examining multiple properties of an instrument rather than treating a single statistic as proof of validity.

Measurement invariance strengthens comparisons across academic discipline

Another useful finding was support for measurement invariance across medical and non-medical students. Measurement invariance asks whether a scale operates sufficiently similarly across groups for score comparisons to be meaningful.

This matters because an apparent difference between two student groups could otherwise be partly produced by the questionnaire functioning differently in each group. Evidence supporting invariance gives greater confidence that observed score comparisons are not simply artifacts of a fundamentally different measurement structure across academic disciplines.

However, measurement invariance does not eliminate the other limitations of observational research. It does not prove causation, establish national prevalence or show that every unmeasured characteristic was balanced between groups.

What universities can learn from overlapping distress

The practical message extends beyond psychometrics. When depression, anxiety and stress are tightly interconnected, mental-health support systems built around only one symptom category may overlook how students actually experience distress.

A student may present because of examination stress but also have severe anxiety, persistent low mood or reduced functioning. Another may seek help for depressive symptoms while significant tension and worry remain unrecognized.

Universities can therefore benefit from mental-health pathways that allow assessment of multiple symptom domains rather than requiring students to identify the correct psychological label before receiving support.

This could include confidential screening pathways, counseling access, referral systems, psychoeducation, sleep-health initiatives and procedures for responding to students whose symptoms interfere substantially with daily functioning. The goal should not be to medicalize ordinary academic pressure, but to make appropriate support available when distress becomes persistent, severe or disabling.

The findings should not be interpreted as prevalence estimates for Pakistan

The study involved 602 undergraduate students from universities in Lahore. Although the sample allowed detailed psychometric analysis, it should not be treated as nationally representative of all Pakistani university students.

Pakistan has substantial diversity in educational systems, socioeconomic circumstances, geographic regions, university cultures and access to mental-health care. Results from one urban student sample may not generalize directly to students elsewhere in the country.

The cross-sectional design also prevents conclusions about how symptoms develop over time. We cannot determine from these data whether stress preceded anxiety, anxiety preceded depression, or whether shared external factors contributed simultaneously to all three.

Self-report questionnaires introduce additional limitations because responses depend on participants’ perceptions, understanding and willingness to report symptoms. Clinical interviews would be required for diagnostic assessment.

A broader lesson for mental-health measurement

One of the most useful lessons from the DASS-21 findings is methodological: good psychometric performance is multidimensional.

A scale can be reliable without its domains being completely distinct. A factor model can fit adequately while its factors remain extremely highly correlated. Group comparisons can be statistically significant while explaining only a small fraction of individual variation.

These are not contradictions. They describe different properties of the data.

For researchers, reporting these details produces a more realistic account of what a questionnaire can and cannot tell us. For clinicians and universities, the findings reinforce the importance of interpreting symptom scores within a wider assessment rather than allowing questionnaire labels to define a person’s mental-health experience.

What our DASS-21 analysis ultimately revealed

Our study supported the reliability and conventional three-factor organization of the DASS-21 among the Pakistani university students surveyed. At the same time, the depression, anxiety and stress factors were exceptionally strongly correlated, particularly anxiety and stress.

This combination is more informative than either result alone. It suggests that the DASS-21 can organize student distress into meaningful domains while also reminding us that the boundaries between those domains may be highly permeable.

The broader implication is not that depression, anxiety and stress should be collapsed into a single clinical concept. It is that student mental health should be understood as interconnected. Researchers should examine discriminant validity carefully, and practitioners should avoid interpreting any single subscale without considering the wider pattern of symptoms and functioning.

Medical disclaimer: This article is for educational and research communication purposes only. DASS-21 scores and other self-report screening results do not establish a psychiatric diagnosis and should not replace assessment by an appropriately qualified healthcare professional. Anyone experiencing persistent, severe or functionally impairing psychological symptoms should seek appropriate professional evaluation. Urgent mental-health concerns require prompt assessment through appropriate local healthcare or emergency services.