Overall, 63% (n = 1524) of the patients were diagnosed with a CMHD (bipolar, depressive, PTSD, anxiety, or somatization disorder). Further, 2% (n = 55) were registered with a mixed anxiety depressive disorder without a comorbid CMHD. Other frequent prevalent diagnoses for patients with no CMHD were attention deficit hyperactivity disorder (10%, n = 245) and personality disorders (4%, n = 106), and symptoms and signs involving emotional state (R45, 21%, n = 496). One year’s prevalence of comorbid CMHD was the following: for patients diagnosed with bipolar disorder (n = 125), 25% also had a depressive disorder, 8% had PTSD, 17% had an anxiety disorder, and 6% had somatization disorder. Out of patients diagnosed with a depressive disorder (n = 865), 4% were also diagnosed with bipolar disorder, 6% with PTSD, 28% with anxiety disorder, and 3% with somatization disorder. Regarding PTSD (n = 245), 4% were diagnosed with bipolar disorder, 23% with depressive disorder, 9% with anxiety disorder, and 2% a somatization disorder. Regarding anxiety disorders (n = 613), 3% had bipolar disorder, 39% had depressive disorder, 3% had PTSD, and 5% had somatization disorder. Of patients with somatization disorder (n = 92), 8% had bipolar disorder, 30% had a depressive disorder, 7% had PTSD, and 32% had an anxiety disorder.
Overall, 22% of patients diagnosed with a CMHD had a comorbid mood and anxiety disorder. Correspondingly, 14% of all patients with diagnostic data had a comorbid mood and anxiety disorder. Patients with comorbid disorders scored higher symptom severity of depression and anxiety than patients diagnosed with non-comorbid disorders, with the only exception of symptoms of anxiety for somatization disorders (Table 1).
Insert Table 1 here
Patients with bipolar disorders had a mean of 4.37 (SD = 3.65) assessments, and 11.23 (SD = 9.71) psychotherapy appointments. For depressive disorders, these numbers were 3.86 (SD = 4.10), and 11.60 (SD = 8.78) respectively. Patients with anxiety disorders had 3.72 (SD = 3.91) assessments and 10.64 (SD = 8.10) psychotherapy appointments, while patients with PTSD had 3.89 (SD = 4.54) assessments and 12.40 (SD = 10.19) psychotherapy appointments. Patients with somatization disorders had 4.45 (SD = 4.26) assessments and 10.62 (SD = 7.34) psychotherapy appointments, and mixed anxiety and depression had 2.69 (SD = 2.47) assessments and 5.04 (SD = 5.98) psychotherapy appointments. Patients with comorbid disorders had a mean number of 4.30 (SD = 4.26) assessments and 13.41 (SD = 9.94) psychotherapy appointments.
Compared to non-comorbid disorders, many comorbid disorders had statistically significant more assessments (bipolar t = 2.16, p = .033; depression t = 2.16, p = .031; anxiety t = 2.59, p = .010) and psychotherapy appointments (bipolar t = 3.41, p < .001; depression t = 4.57, p < .001; anxiety t = 7.28, p < .001; PTSD t = 2.15, p = .032). Between non-comorbid CMHD, the statistically significant differences was between psychotherapy appointments for depression (M = 10.59, SD = 7.83) and anxiety (M = 8.71, SD = 6.21; t = 3.84, p < .001), between PTSD (M = 11.59, SD = 9.04) and bipolar (M = 9.36, SD = 7.83; t = 1.98, p = .049), and between PTSD and anxiety (t = 4.34, p < .001).
Patients with mixed anxiety and depression had significantly fewer assessments (bipolar t = 3.10, p = .002; depression t = 2.09, p = .037; somatization t = 2.78, p = .001) and psychotherapy appointments (bipolar t = 4.38, p < .001; depression t = 5.46, p < .001; anxiety t = 5.00, p < .001; PTSD t = 5.24, p < .001; somatization t = 4.77, p < .001) compared to many of the other CMHD.
Factor structure results
The latent three-dimensional structure of cognitive-, somatic depression and anxiety freely estimating two unique covariances reached acceptable model fit (χ2 = 872.436, df = 85, p < .001; SRMR = .046; RMSEA = .061 [90% CI = 0.058, 0.065]; CFI = .940; TLI = .926).
Latent Class Analysis results
Model fit indices for the LCA are presented in Table 2. The 1-Class model had the largest AIC, BIC, and aBIC, thus demonstrating the worst model fit. The LMR test, aLMR test, and BLRT in the 2-Class model solution all had p values < .01, indicating to reject the 1-Class model solution in favour of a 2-Class model solution. Statistically significant p values for the LMR and BLRT indicated that the current (k-class) model fitted the data better than the model with one less class (k-1 class). Results from comparing the 3-Class to the 2-Class model solution favoured a 3-Class model solution, which had lower criterion indices than the 2-Class model solution. Similarly, the 4-Class model and 5-Class model all had smaller criterion indices. Although the criterion fit indices showed that there was an improvement in model fit when comparing the 3-Class model solution to the 4-, and 5-Class model solutions, the deterioration in the entropy fit statistic for the 4-Class model solution was more pronounced, followed by the 5-Class model solution, which indicates that the 4- and 5-Class model solutions contain classes that are not clearly separated. Higher entropy values indicate that classes are easily distinguishable and distinctive, and as such favoured the 3-Class model solution which had a relatively high entropy value. The three classes were labelled as; high distress class (43%), moderate distress class (41%), and low distress class (16%) since they only differed by the degree of symptom severity.
Factor Mixture Model results
Results from the factor mixture model analysis are presented in the bottom part of Table 2. The FMM with the lowest criterion indices was the three-factor, 4-Class FMM-2 model. However, one of the classes in this model solution turned out to be spuriously extracted from the data as it contained no respondents, so we examined the next lowest criterion indices —the three-factor, 5-Class FMM-2 model. The LMR test, aLMR test, and BLRT in the three-factor, 5-Class FMM-2 model solution all had p values greater than .05, which indicates that this model solution should be rejected despite lower criterion indices. Therefore, we then considered a three-factor, 3-Class FMM-2 model which had the second lowest criterion indices. Although the LMR and the aLMR tests did not show univocally support for this model solution, the BLRT had a p-value < .001, indicating that this model solution significantly fit the data. The entropy value for this model solution was also high, indicating that there is a clear separation between distinguishable classes. Furthermore, the three-factor, 3-Class FMM-2 model replicates the combined results from the CFA and LCA analyses.
Insert Table 2 here
Selecting the three-factor, 3-Class FMM-2 model implies that the underlying symptoms are conceptualized equivalently and normally distributed within classes. In other words, the three classes are represented by normally distributed patterns of symptoms of depression and anxiety such that individuals within classes can have quantitatively different ranges of symptom severity. The criterion indices for the FMM-1 solution were much higher and therefore unsuitable for model selection. Additionally, the assumptions of the FMM-1 imply that all patients within a class are having the same levels of distress and that there is no within-class heterogeneity in symptoms of depression and anxiety. This is unlikely to be correct as symptom variation exists as well as the range of severity, consistent with our FMM-2 model solution. Since we assumed variations in symptoms of depression and anxiety as well as the range of severity, the three-factor, 3-Class FMM-2 model was chosen.
Interpretation of classes from the factor mixture model
The three distress classes were labelled according to differences in factor means (see Figure 1). The reference Class two comprising 40% of the sample, was labelled “mixed depression and anxiety” whereas patients in Class one (33%), in comparison to the mixed depression and anxiety class reported lower levels of cognitive depression, but higher levels of somatic depression and anxiety symptoms and was thus labelled “anxiety and somatic depression”. Patients in Class three (27%) on the other hand reported lower somatic depression than mixed depression and anxiety class and Class three was thus labelled “cognitive depression”. The factor variances within the classes were all significant, which agrees with the interpretation that there are variations in diagnostic class membership and the range of severity of the patient’s self-reported symptoms of depression and anxiety.
Insert Figure 1 here
Predictors of class membership
See Table 3 for predictors of class membership. There were no findings of gender predicting any class membership. Older age predicted a higher probability of membership to the anxiety and somatic depression class, compared to both the mixed depression and anxiety and the cognitive depression classes. Being single predicted a higher probability of membership in the anxiety and somatic depression class compared to the mixed depression and anxiety class, and the cognitive depression class compared to the mixed depression and anxiety class.
Insert Table 3 here
Differences in relevant outcome variables
Patients with a high probability of membership in the anxiety and somatic depression class were to a larger degree associated with being on sick leave and being diagnosed with depression, anxiety and comorbid mood/anxiety disorder compared to the other classes.
Patients with a high probability of belonging to the cognitive depression class were to a larger degree associated with being on sick leave and being diagnosed with depression, and comorbid mood/anxiety disorder, compared to the mixed depression and anxiety class. They also had a higher probability of being diagnosed with mixed anxiety and depressive disorder, compared to the anxiety and somatic depression class. Regarding the mixed depression and anxiety class, they had a higher probability of being diagnosed with somatization disorder compared to the cognitive depression class. Regarding functional impairment and service use, the anxiety and somatic depression class had more assessment and psychotherapy appointments and reported the highest degree of impairment compared to the other classes. The cognitive depression class reported more psychotherapy appointments and functional impairment than the mixed depression and anxiety class. Outcomes across classes are presented in Table 4.
Insert Table 4 here