Data were collected in a naturalistic observational study that ran from May 2017 – March 2020 at the Department of Mental Health and Substance Abuse, Diakonhjemmet Hospital in Oslo, Norway. The clinic is part of the national health service, and the study is part of the project “The Norwegian studies of psychological treatments and work (NOR-WORK)”. Patients are referred by their general practitioners for treatment of depression and anxiety. Patients at the clinic are generally of working age, and previous research has shown that on average, half the patients are on sick leave due to depression or anxiety at baseline [22]. They are then screened by a clinical psychologist using anamnestic information, the Beck Depression Inventory-II (BDI-II), the Beck Anxiety Inventory (BAI), and the MINI-International Neuropsychiatric Interview [23,24,25]. Patients are diagnosed during the screening in accordance with the International Classification of Diseases 10 (ICD-10) [26]. Inclusion criteria for the present study were that the patient was an adult of working age (18–70 years) with clinically significant levels of depression and anxiety operationalised as follows: Patients with a primary depression diagnosis had to have a minimum score of 14 on the BDI-II, and patients with a primary anxiety diagnosis had to have a minimum score of 16 on the Beck Anxiety Inventory BAI. In addition to primary depression or anxiety diagnoses, patients with adjustment disorder and mixed anxiety and depression were included in the study. Adjustment disorder is sometimes referred to as “situational depression”, underlining its close relationship with depressive disorders [26]. Similarly, patients with a mixed anxiety and depressive disorder were included as the diagnosis is comprised of symptoms of anxiety and depression.

Exclusion criteria were severe mental illness such as bipolar disorder, high risk of suicide, engaging in active substance abuse, or suffering from cluster A or B personality disorder. Patients scoring below clinical thresholds for depression and anxiety on the BDI and BAI at baseline were excluded from the study. All patients who signed a written consent form and completed treatment, including filling in questionnaires at baseline and at end of treatment, were included (N = 416). The current study thus focused on patients who completed treatment.

Patients received either Metacognitive therapy (MCT) or Cognitive behavioural therapy (CBT) according to diagnose-specific manuals [27, 28], and average duration of treatment was 10.11 sessions (SD 3.93). Previous research has shown that half the patients are on sick leave when referred, and treatment thus also includes interventions aimed at helping patients return to work [29].

Clinical and sociodemographic data were collected at baseline and end of treatment from patient journals and from self-report questionnaires.

The EQ-5D-5L: The EQ-5D-5L questionnaire firstly asks respondents to rate their current health on five dimensions: Mobility, Self-Care, Usual activities, Pain / discomfort, and Anxiety / depression on a severity scale from 1 (“No problems”) to 5 (“Severe problems”). The combined severity ratings give an EQ-5D profile, e.g. “11111” in the case of “No problem” on all five dimensions. This health profile can be converted to the EQ-5D value using preference-based weights. A value of 0.00 indicates death and 1.00 indicates perfect health. The EQ-5D value can be used to calculate quality-adjusted life-years (QALYs), i.e. a score of 1.00 for one year equals one QALY. The preference-based weights used to convert responses to EQ-5D values are often referred to as “value sets”. A study is underway, but there is currently no Norwegian value set [30]. This study used the crosswalk system recommended by NICE for converting EQ-5D profiles to EQ-5D values [31, 32]. For the EQ-5D value, healthy people generally report scores close to 1.0. In a recent survey of the Norwegian general population, the mean EQ-5D value in a postal survey was 0.848 [33].

The second part of the EQ-5D-5L asks patients to rate their health on a 20 cm visual analogue scale (VAS) where the bottom (“0”) indicates worst imaginable health, and the top (“100”) indicates best imaginable health. Although it is related to the EQ-5D profile and the value scores, it does not measure the same construct. For instance, the EQ VAS score has been shown to decline with age even for people whose EQ-5D profile show no problems (“11111”) [8].

The Beck Depression Inventory-II (BDI-II) is a 21-item questionnaire measuring severity of symptoms over the last two weeks on a scale from 0 to 3, giving a total sum score of 0–63. Examples include feeling sad and change in appetite or sleep. Suggested scoring indicates that 0–13 reflects minimal symptoms, 14–19 mild, 20–28, moderate, and 29–63 severe symptoms [24]. The BDI-II has been found to be psychometrically sound in depression[31], Chronbach’s α in the current study was 0.86.

The Beck Anxiety Inventory (BAI) is a self-report measure of anxiety severity over the last week. As with the BDI-II, anxiety symptoms (e.g. “Heart pounding or racing” or feeling “nervous”) are scored on a severity range from 0 to 3, giving a total sum score of 0–63. Suggested scoring indicates that 0–15 reflects mild symptoms, 16–25 moderate, and 26–63 severe symptoms. The BAI has demonstrated good psychometric properties [34], Chronbach’s α in the current study was 0.90.

Descriptive statistics on age, gender, education level and diagnosis were compiled at baseline. Distribution of scores on the EQ-5D dimensions were calculated in percentages at baseline and at end of treatment and analysed using a non-parametric test of trends developed by Cuzick. The test is similar to the Wilcoxon rank-sum test [35]. Mean scores and standard deviations at baseline and end of treatment, including change (∆) during treatment, were calculated for the BAI, the BDI-II, the EQ-5D values, and the EQ VAS. Effect sizes (ES) were calculated from baseline to end of treatment using Cohen’s d. Values < 0.5 are considered small, ≥ 0.5 < 0.8 moderate, and ≥ 0.8 large [36]. We also calculated the standardised response mean (SRM), defined as the mean change in score from baseline to end of treatment divided by the standard deviation of change in scores [37]. For the SRM it is suggested that magnitude of change is dependent on correlation between scores at baseline and end of treatment. For example, SRM > 0.8 can be interpreted as large if this correlation < 0.5, moderate if correlation > 0.5 [38]. Agreement between the change scores on the four measures were also analysed with Pearson’s correlation. Pearson’s correlation < 0.40 are considered weak, 0.40–0.49 moderate, and > 0.50 are considered strong [39].

Using the BAI and the BDI-II, the patients were then classified according to treatment response. With a minimum score of 14 on the BDI-II for depression patients and 16 on the BAI for anxiety patients at baseline, based on scoring norms for the BDI-II and BAI, patients were classified thus: “Deteriorated” if their scores increased by 9 points or more from baseline to end of treatment, “Unchanged” if the change was less than 9 points in either direction, and “Improved” if the scores decreased by 9 points or more but score at the end of treatment was still above the clinical threshold, . Finally, patients were classified as “Recovered” if their score decreased by 9 points or more and their final score was below clinical threshold (i.e. 14 for the BDI-II and 16 for the BAI) [18, 40, 41].

We ran ROC curve analyses to determine how well the EQ-5D value scores could correctly classify patients according to the clinical criteria of the BDI-II and the BAI: Recovered versus Improved, Recovered versus Unchanged, and Improved versus Unchanged. Analyses of BDI were run to calculate the area under the curve (AUROC) using the entire sample for patients that had a BDI score of at least 14 at baseline, and for all patients who had a BAI score of at least 16, regardless of primary diagnoses. Then, using primary diagnosis as recorded from the medical journals, we then calculated the AUROC for BDI-II for only the patients with depression as primary diagnosis and BDI-II baseline scores of at least 14. Lastly, we calculated the AUROC for BAI for the patients with anxiety as their primary diagnosis and a BAI baseline score of at least 16. The EQ-5D value at end of treatment was used as classifier, when computing the AUROC. AUROC was interpreted as < 0.50 useless test, 0.51–0.69 poor test, 0.7–0.79 fair test, 0.8–0.89 good test, 0.9–0.99 excellent test, 1.0 perfect test [40]. We calculated the sample size needed for the groups included in the ROC analyses. We set the Alpha level to 0.05 and the Beta level to 0.20, area under curve was set to 0.7 and value of null hypothesis was set to 0.5. The ratio of positive to negative cases was set according to the characteristics of the sample. We also computed cut-off values for recovery using Youden’s index (J), which displays which values have the highest combined sensitivity and specificity [42].

Generally accepted methods for handling missing data are applicable to the EQ-5D-5L [8]. Missing data on individual items in the current study were replaced by weighted means, a method developed for treating missing data in depression cohorts [43]. All analyses were carried out using STATA 16 [44].

All patients included in the study gave written, informed consent to participate. The study is classified as health service research under Norwegian regulation. The Norwegian Data Protection Agency has in such cases designated that treatment providers (i.e. hospitals) are responsible for proper data management. Data collection and security in the present study was managed by Diakonhjemmet Hospital, and approval of data handling was granted by Oslo University Hospital, approval number 2015/15606. The study was carried out in accordance with the principles of the Helsinki declaration.