Mood and anxiety disorders are often comorbid 12. This finding is not surprising given the symptom overlap between these two syndromes 3. A Tripartite model of Affect has been proposed to account for this overlap and to identify the unique components of anxiety and depression 4. Under this model, Negative Affect (NA), a mixed anxiety-depression factor, explains the comorbidity between these syndromes. Positive Affect (PA) is the depression-specific factor and Physiological Hyperarousal (PH) is unique to anxiety.

There is some evidence for the Tripartite model for its ability to distinguish between depression and anxiety 56789. However, some findings have suggested the need to modify the original model 10. For example, PH may only be related to Panic Disorder and, to a lesser extent, Generalised Anxiety Disorder.

Much of the research into the validity of the tripartite model has been conducted with measures that were not specifically designed to measure the tripartite constructs (e.g. BDI, BAI). Specific measures may be necessary to explore the utility of the model 11.

The Mood and Anxiety Symptom Questionnaire (MASQ) was designed to be a specific measure of the Tripartite model of Affect 1213. Three general distress scales measure the postulated shared factor, NA. A depression- (Anhedonic Depression, AD) and anxiety-specific (Anxious Arousal, AA) scale measures PA and PH respectively. Watson et al. reported good discriminant and convergent validity for the MASQ across five samples comprising college students, adults sampled from the community and patients presenting to a pain clinic. Factorial validity for the MASQ has been established in non-clinical samples, with three factors consistently found to best represent the data 71314. However, a recent study employed confirmatory factor analysis with depressed and anxious patients and reported that the structure of the tripartite model was not confirmed in this clinical sample 15. Additionally, two recent studies have questioned the discriminant validity of the MASQ in clinical samples 1617. Discriminant validity reflects the extent to which an instrument measures the construct it is purported to measure, without inadvertently measuring other domains/constructs. Buckby et al. reported that depressed participants rated higher than non-depressed participants on all MASQ scales, including the anxiety-specific AA. Furthermore, they reported that participants with anxiety disorders did not score higher than participants without a current disorder on any scale. Boschen and Oei 16 reported that although the MASQ scales significantly discriminated between anxiety and depression, there was little clinical utility in these findings as the maximum discrimination was only 70%. Both studies cautioned future researchers to be mindful of the potential weakness of the MASQ's validity in clinical samples and indicated that further research, including replication, was necessary to better understand this issue.

Boschen & Oei 16 used a clinical sample of adults with Mood and Anxiety disorders. A recent review has criticised researchers for unduly focusing on psychometric characteristics and neglecting sensitivity 18. Using similar analyses to Boschen & Oei, the present study therefore attempted to replicate their findings in a clinical sample of adolescents and young adults in order to determine how generalisable those findings were. People in the early stages of mental disorders may be especially receptive to treatment and may require more benign interventions than chronic populations 19. It is therefore important to be able to validly recognise psychological distress in people at-risk of mental disorder. Should the MASQ be determined to validly detect depression and anxiety, it could prove to be a useful tool that is easily administered by professions including GPs and triage workers. It was hypothesised that the MASQ would show marginal to weak criterion validity in the present sample. A secondary aim was to compare the MASQ depression scale, AD, with the Center for Epidemiologic Studies Depression Scale (CES-D) 20. The CES-D is a widely-used measure of depressive symptomatology that has been validated in both adult and adolescent samples 2021. However, some researchers have questioned its specificity to depression. For example, the CES-D has high reported correlations with anxiety measures and contains items that are not specific to depression. The present study sought to determine whether the MASQ offered an improvement to this earlier measure.

Participants were 150 people aged 15–24 years consecutively referred to ORYGEN Youth Health (OYH), a specialist youth public mental health service in Melbourne, Australia. Sources of referral included family, GPs, hospitals and school counsellors.

Between April and September, 2003, all young people who were referred to OYH with non-psychotic disorders were approached by trained research interviewers. Consenting participants completed the MASQ and a diagnostic interview. Inter-rater assessments were conducted in approximately 15% of the cases to ensure agreement between interviewers. Kappa values for mood (0.89) and anxiety diagnoses (0.80) were high.

The sample, setting and methodology have been described in greater detail elsewhere 2223.

Two hundred and four people met the study criteria during the recruiting phase. Of the 150 consenting participants 14 did not return valid questionnaires, leaving a sample of 136 with useable data. There was no significant demographic differences between those with, and without missing data. A small proportion of participants (< 10%) had several, random missing variables replaced with the Expectation Maximisation method 25.

The mean age of participants was 18.11 years (SD = 2.61) and the sample was comprised of 61% females (N = 83). There was no significant gender difference in MASQ scores. There was a significant difference between adolescents (age 15–17) and young adults (age 18–24) on all MASQ scales except AA with older participants scoring higher.

The majority of the sample (80.1%, N = 109) rated for a current Axis-I disorder. Mood and Anxiety disorders were most common. The sample was split according to diagnostic status: Mood Disorder only (no comorbid anxiety, N = 29, 21.3%); Anxiety Disorder only (no comorbid mood, n = 22, 16.2%); Comorbid Anxiety-Mood (n = 35, 25.7%); and no Anxiety or Depression (n = 50, 36.8%). The most common Anxiety disorders were specific phobia (16.9%), social phobia (15.4%) and panic disorder (11.0%). Post traumatic stress (8.1%), generalised anxiety (5.1%) and obsessive compulsive (6.6%) disorders were less frequent. Nearly half (46%) of the no anxiety-mood group rated for another Axis-I diagnosis (including substance abuse/dependence and eating disorders). Prevalence of other disorders has been described elsewhere 2.

Means and standard deviations are presented in Table 1. Inter-correlations between all MASQ scales were uniformly high and statistically significant with the lowest correlation between the disorder-specific scales AA and AD (r = 0.59, see Table 2). Inter-correlations between AA:AD were inspected across the diagnostic groups. This correlation was high in all groups rating for either a Mood or Anxiety Disorder with the highest correlation in Anxiety only (r = 0.62), followed by Comorbid (r = 0.60) and Mood only (r = 0.46) (p < .001 for all). In contrast, AA:AD was only moderately correlated in the no Anxiety/Mood group (r = 0.35, p = 0.01). When examining those participants without any current diagnoses (N = 27), AA:AD were uncorrelated (r = 0.18, p = 0.36). This finding indicated a differential relationship for the correlation between those with (r = 0.59, p < .001, N = 109) and those without an Axis-I disorder. Therefore, an independent samples t-test of the correlational coefficients was conducted 26, showing a significantly higher correlation in a composite diagnosis group that included participants with any current disorder (z = 4.04, p < .001).

To replicate earlier findings 16 logistic regression analyses were run with diagnostic status as the dependent variable. In the first analysis, a dichotomous outcome variable was created reflecting presence (n = 64, 47.1%) versus absence (n = 72) of a Mood Disorder. The depression-specific scale AD was included in the first block of covariates, the three GD scales in the second block and AA in the third block.

AD significantly predicted mood disorder (χ² (1) = 44.06, p < .001). The addition of the GD scales in the second block did not significantly increase the prediction of mood disorder (p = 0.09), nor did inclusion of AA at the third step (p = 0.78). The overall predictive accuracy of AD for a mood disorder was 72.8% (71.9% for presence and 73.6% for absence of a mood disorder).

In the second analysis, presence (n = 57, 41.9%) versus absence (n = 79) of an anxiety disorder was the dependent variable. The anxiety-specific scale AA was included in the first step, the GD scales in the second, and AD in the third step.

AA significantly predicted anxiety disorder (χ² (1) = 4.58, p = 0.03). The GD scales did not increase the prediction (p = 0.15), nor did AD in the final step (p = 0.79). The overall predictive accuracy of AA for an anxiety disorder was 61.0% (29.8% for presence and 83.5% for absence).

The ROC curve allows researchers to graphically determine a test's sensitivity and specificity 27. The area under the curve of a ROC plot provides a score that ranges from 1 (test is always correct) to 0 (test is never correct). A score of 0.5 indicates that the result is no better than chance. When comparing different tests, the one with the larger area under the curve is determined to be the better diagnostic tool.

In the first analysis, the dichotomous Mood disorder variable was entered as the state variable and the disorder-specific scales AA and AD and the CES-D were entered as the test variables. The ROC plot for AD contained 81.8% under the curve (SE = 0.04, 95% C.I. = 0.75, 0.89), AA contained 72.3% (SE = 0.04, 95% C.I. = 0.64, 0.81) and CES-D contained 78.7% (SE = 0.04, 95% C.I. = 0.71, 0.86).

In the second analysis, the dichotomous anxiety variable was input as the state variable and the disorder-specific scales as the test variables. The ROC plot for AA contained 61.7% under the curve (SE = 0.05, 95% C.I. = 0.52, 0.72), AD contained 62.0% (SE = 0.05, 95% C.I. = 0.53, 0.72) and CES-D contained 58.3% (SE = 0.05, 95% C.I. = 0.48, 0.68).

The sensitivity and specificity of AD was further inspected to identify a potential cut-off to indicate 'caseness' for depression. By visually inspecting the ROC for the largest area under the curve and by manually comparing various true positive and true negative rates, it was determined that an AD cut-off of 76 best reflected caseness (sensitivity = 85%, specificity = 65%). These figures were then compared to CES-D (cut-off = 24). AD is equivalent to, if not better than, the CES-D (sensitivity = 84%, specificity = 61%) in discriminating Mood Disorders when utilising the proposed cut-off of 76. Further analyses with AA were not run as this scale demonstrated poor discriminant validity (see above) (Figures 1 and 2).