Life satisfaction in persons with mental disorders

Clinical records of inpatients treated at the Schoen Clinic Roseneck (Prien am Chiemsee, Germany) between 2014 and 2019 were analyzed. The German version of the SWLS [5] and the depression scale of the Patient Health Questionnaire (PHQ–9 [14‐16]) are part of the routine diagnostic assessment at the hospital and are completed by the patients both at admission and at discharge. Only patients without missing SWLS data at admission and discharge were included in the current analyses. Moreover, only data from diagnostic groups with n > 300 were included as smaller group sizes are generally considered as not appropriate for confirmatory factor analysis ([17]; the largest group of patients that was excluded because of this procedure was other anxiety disorders [F41; n = 164]). The final sample with complete SWLS data at both admission and discharge was N = 9649 patients (n = 9610 for PHQ–9 data) and included patients of seven diagnostic categories (based on ICD–10 classification): depressive episode, recurrent depressive disorder, phobic disorders, obsessive–compulsive disorder, trauma-related disorders, somatoform disorders, and eating disorders (Table 1).

Sample characteristics. Groups were compared regarding sex distribution with a χ²-test and regarding age and length of stay with univariate analyses of variance using IBM SPSS Statistics version 24.

Internal reliability, factor structure, and measurement invariance. Internal reliability of the SWLS was evaluated with McDonald’s ω (as has been recommended [18‐21]), which was calculated with JASP version 0.11.1 (https://​jasp-stats.​org [22]). Confirmatory factor analysis was conducted with the structural equation modeling module of JASP, which is based on the R-package lavaan (https://​lavaan.​ugent.​be). Diagonally Weighted Least Squares was chosen as estimation method because of the ordinal scale structure [23]. In line with previous studies [4, 5], a one-factor model was specified. Model fit was considered as good according to the recommendations by Schermelleh-Engel et al. [17]: Comparative Fit Index (CFI) ≥ 0.97, Goodness of Fit Index (GFI) ≥ 0.95, Root Mean Square Error of Approximation (RMSEA) ≤ 0.05, and Standardized Root Mean Square Residual (SRMR) ≤ 0.05. Measurement invariance across groups was tested at four levels: configural invariance (tests if the configuration of the model is the same across groups), metric invariance (tests if the factor loadings are the same across groups), scalar invariance (tests if the intercepts are the same across groups), and strict invariance (tests if the residual variances are the same across groups). There are different recommendations of how to evaluate measurement invariance but a fairly well established guideline is that model fit changes of ΔCFI ≤ 0.01 indicate invariance [24, 25]. We do not report the χ²-test of exact fit or χ²-difference tests between models because these are usually significant in large samples and, therefore, uninformative in the current sample. All analyses on internal reliability, factor structure, and measurement invariance were run separately for SWLS scores at admission and discharge.

Changes in life satisfaction as a function of sex, age, and length of stay. Three linear regression analyses were calculated separately with sex, age, and length of stay as independent variable with PROCESS version 3.4 (https://​processmacro.​org [26]). Group was entered as multicategorical moderator variable using indicator coding [27] and changes in SWLS scores (SWLS scores at discharge minus SWLS scores at admission) were entered as dependent variable.

Associations between life satisfaction and depressive symptoms. Pearson correlation coefficients were calculated using IBM SPSS Statistics version 24 to examine associations between SWLS and PHQ–9 scores at admission and discharge. To examine the relationship between changes in SWLS scores from admission to discharge and changes in PHQ–9 scores from admission to discharge, a repeated measures correlation was calculated with the R-package rmcorr [28].

Groups differed in sex distribution (χ²₍₆₎ = 1302, p < 0.001, ϕ = 0.367) with patients with phobic disorders having the highest percentage of males and patients with eating disorders having the lowest percentage of males (Table 1). Groups also differed in age (F_(6,9642) = 744, p < 0.001, η²_p = 0.316) with patients with recurrent depressive disorder having the oldest age and patients with eating disorders having the youngest age (Table 1). They also differed in length of stay (F_(6,9642) = 556, p < 0.001, η²_p = 0.257) with patients with somatoform disorders staying the shortest and patients with eating disorders staying the longest (Table 1).

Admission. Internal reliability was good (McDonald’s ω = 0.861). The one-factor model showed a good fit (CFI = 0.997, GFI = 1.00, RMSEA = 0.043, SRMR = 0.023). Factor loadings are displayed in Fig. 1a. Examination of model fit changes indicated configural invariance (ΔCFI = 0.000 compared to the baseline model), metric invariance (ΔCFI = 0.001 compared to the configural model), scalar invariance (ΔCFI = 0.004 compared to the metric model), and strict invariance (ΔCFI = 0.002 compared to the scalar model). Accordingly, the strict invariance model still showed good model fit (CFI = 0.990, GFI = 0.998, RMSEA = 0.041, SRMR = 0.044).

Discharge. Internal reliability was good (McDonald’s ω = 0.897). The one-factor model showed a good fit (CFI = 0.999, GFI = 0.999, RMSEA = 0.031, SRMR = 0.019). Factor loadings are displayed in Fig. 1b. Examination of model fit changes indicated configural invariance (ΔCFI = 0.001 compared to the baseline model), metric invariance (ΔCFI = 0.002 compared to the configural model), scalar invariance (ΔCFI = 0.004 compared to the metric model), and strict invariance (ΔCFI = 0.001 compared to the scalar model). Accordingly, the strict invariance model still showed good model fit (CFI = 0.993, GFI = 0.999, RMSEA = 0.043, SRMR = 0.045).

A main effect of group (F_(6,9642) = 55.2, p < 0.001, η²_p = 0.033) indicated that diagnostic groups differed in life satisfaction. Patients with somatoform disorders reported the highest life satisfaction (M = 19.4, SE = 0.34) and patients with trauma-related disorders reported the lowest life satisfaction (M = 14.4, SE = 0.27; Fig. 2). A main effect of time (F_(1,9642) = 1388, p < 0.001, η²_p = 0.126) indicated that life satisfaction increased from admission (M = 15.9, SE = 0.09) to discharge (M = 18.5, SE = 0.10) across diagnostic groups (Fig. 2). These main effects, however, were qualified by a significant interaction of group × time (F_(6,9642) = 8.33, p < 0.001, η²_p = 0.005). Patients with somatoform disorders had the smallest increase in life satisfaction (M = 1.80, SE = 0.27) and patients with eating disorders had the largest increase in life satisfaction (M = 3.32, SE = 0.10). At admission, mean SWLS scores were below the neutral score of 20 [1, 4] in all groups. At discharge, mean SWLS scores were above a score of 20 only in patients with a depressive episode (M = 20.4, SE = 0.16) and patients with somatoform disorders (M = 20.3, SE = 0.38).

Sex. The interaction of sex × group was not significant (R² change = 0.001, F_(6,9635) = 1.00, p = 0.425).

Age. The interaction of age × group was significant (R² change = 0.002, F_(6,9635) = 3.32, p = 0.003). A younger age was significantly related to larger increases in life satisfaction in patients with a depressive episode (b = –0.02, SE = 0.01, p = 0.005, r = –0.064), recurrent depressive disorder (b = –0.03, SE = 0.01, p < 0.001; r = –0.075), and trauma-related disorders (b = –0.04, SE = 0.02, p = 0.007; r = –0.110), and marginal significantly related to larger increases in life satisfaction in patients with obsessive–compulsive disorder (b = –0.02, SE = 0.01, p = 0.054; r = –0.065) and somatoform disorders (b = –0.03, SE = 0.02, p = 0.084; r = –0.098). Age was not associated with changes in life satisfaction in patients with phobic disorders and eating disorders (both ps > 0.147; Fig. 3).

Length of stay. The interaction of length of stay × group was significant (R² change = 0.002, F_(6,9635) = 3.77, p = 0.001). A longer stay was significantly related to larger increases in life satisfaction in patients with trauma-related disorders (b = 0.02, SE = 0.01, p = 0.014, r = 0.102) and marginal significantly related to larger increases in life satisfaction in patients with somatoform disorders (b = 0.03, SE = 0.02, p = 0.079, r = 0.100). A longer stay was significantly related to smaller increases in life satisfaction in patients with recurrent depressive disorder (b = –0.01, SE = 0.01, p = 0.016, r = –0.048) and phobic disorders (b = –0.03, SE = 0.01, p = 0.015, r = –0.118). Length of stay was not associated with changes in life satisfaction in the other groups (all ps > 0.228; Fig. 4).

A main effect of group (F_(6,9603) = 61.2, p < 0.001, η²_p = 0.037) indicated that diagnostic groups differed in depressive symptoms. Patients with obsessive–compulsive disorder had the lowest depression scores (M = 9.60, SE = 0.17) and patients with trauma-related disorders had the highest depression scores (M = 13.8, SE = 0.22). A main effect of time (F_(1,9603) = 4074, p < 0.001, η²_p = 0.298) indicated that depressive symptoms decreased from admission (M = 13.4, SE = 0.08) to discharge (M = 8.60, SE = 0.08) across diagnostic groups. These main effects, however, were qualified by a significant interaction of group × time (F_(6,9603) = 33.7, p < 0.001, η²_p = 0.021). Patients with somatoform disorders had the smallest decrease in depressive symptoms (M = –3.78, SE = 0.27) and patients with eating disorders had the largest decrease in depressive symptoms (M = –6.19, SE = 0.11). At admission, mean PHQ–9 scores were above the cut-off score of 10 [29] in all groups. At discharge, mean PHQ–9 scores were below the cut-off score of 10 in all groups except in the patients with trauma-related disorders (M = 11.6, SE = 0.24).

At admission, SWLS and PHQ–9 scores were moderately, negatively correlated (r = –0.495, p < 0.001). At discharge, SWLS and PHQ–9 scores were highly, negatively correlated (r = –0.610, p < 0.001). The repeated measures correlation between SWLS and PHQ–9 scores within individuals across admission and discharge was r = –0.601 (p < 0.001).

Groups significantly differed in SWLS scores. Patients with trauma-related disorders showed the lowest scores of all groups, starting at 13.3 at admission and increasing to 15.5 at discharge. In line with this, similar scores of 12.9 have been previously reported in a group of individuals with traumatic brain injury and post-traumatic stress disorder [1]. Thus, patients with trauma-related disorders are among the groups with the lowest life satisfaction, with lower scores having only been reported in the literature for male prison inmates and sex workers [1, 4]. Patients with somatoform disorders had the highest scores of 18.5 at admission. Thus, although patients with mental disorders generally report lower life satisfaction than persons without mental disorders, life satisfaction also varies considerably between different diagnostic groups.

Life satisfaction increased during treatment across diagnostic groups. Patients with somatoform disorders showed the smallest increases, which may be explained by the fact that they also had the shortest stay. In line with this, a longer stay (weakly) related to larger increases in life satisfaction in this group. While each group showed a statistically significant increase in life satisfaction during inpatient treatment, only two groups increased beyond the neutral point of 20 of the SWLS [1, 4]: patients with a depressive episode (score of 20.4 at discharge) and somatoform disorders (score of 20.3 at discharge). All other groups still had mean sum scores below 20 at discharge. Thus, although life satisfaction increases during inpatient treatment, most patients are still slightly dissatisfied with their life. However, preliminary data in adolescents with anorexia nervosa from our hospital suggest that life satisfaction actually shows a further increase in the year after discharge [30]. Yet, further studies that examine long-term changes in life satisfaction after inpatient treatment in other patient groups are necessary to corroborate such an effect.

Relationships of age and length of stay with changes in life satisfaction differed across groups. A younger age was predictive of larger increases in life satisfaction, but only in some diagnostic groups such as those with a depressive episode or with recurrent depressive disorder. This is in line with the findings by Meyer et al. [12] who found a negative relationship between the duration of the disorder and life satisfaction in major depression. That is, an older age and a longer duration of the disorder seems to relate to a higher symptom severity and, similarly, to smaller changes in life satisfaction during treatment in depressive disorders. Relationships between length of stay and changes in life satisfaction even showed opposite patterns across groups: A longer stay was associated with larger increases in life satisfaction in some groups while it was inversely (or unrelated) to changes in life satisfaction in other groups. This may suggest that some patients (e.g., those with trauma-related disorders) profit from a longer treatment while in other groups (e.g., those with recurrent depressive disorder) a longer stay may reflect a therapy-resistant course.

In line with previous findings [1, 4], depressive symptoms were negatively correlated with life satisfaction cross-sectionally. In addition, within-person correlational analyses showed that increases in life satisfaction were strongly related to decreases in depressive symptoms. Accordingly, group differences and changes in life satisfaction and depressive symptoms were largely similar. However, they were not identical. For example, the patients with obsessive–compulsive disorder had the lowest depressive symptoms although they did not report the highest life satisfaction. In addition, depressive symptoms showed a clinically significant reduction (i.e., a score lower than 10 on the PHQ–9) in almost all groups while life satisfaction showed an increase above the neutral point of the SWLS in only two groups. That is, a low depression severity is not equivalent to a high life satisfaction and a clinically significant reduction in depression severity does not necessarily imply that one is satisfied with his or her life. Together, these results corroborate that life satisfaction is related to the affective aspects of subjective well-being, but that it is also partially independent from them [1, 4, 31].

Several clinical implications can be derived from the current study. First, this study showed that inpatient treatment of mental disorders not only decreases symptoms such as depression, but also increases life satisfaction. In terms of clinical significance, however, results suggested that treatment reduced depressive symptoms more than it increased life satisfaction. This highlights the need for incorporating other therapy elements in inpatient treatment or in aftercare that are specifically designed to enhance life satisfaction (e.g., [32].). Second, the current study identified subgroups of individuals with mental disorders that may need special attention and in which such targeted interventions for improving life satisfaction might be particularly effective. For example, patients with trauma-related disorders showed the lowest life satisfaction at both admission and discharge and patients with somatoform disorders showed the smallest increase in life satisfaction from admission to discharge. Thus, these diagnostic groups may require a more intensive treatment program that focuses on life satisfaction. Third, the current study also identified moderators of treatment changes in life satisfaction. For example, in both patient groups with trauma-related and somatoform disorders, a longer stay at the hospital related to larger improvements in life satisfaction. This implies that these patients may profit from time-extended treatment.

The following limitations need to be considered when interpreting the current results. First, inpatients with mental disorders may not be representative of the entire population of persons with mental disorders. Specifically, inpatients usually have a higher clinical impairment and distress than outpatients or persons that do not receive treatment. Therefore, SWLS scores reported in the current study should not be treated as norm data for persons with mental disorders as these are likely lower than in persons with mental disorders who are not receiving inpatient treatment. As all patients were treated at the same hospital, site-specific effects can also not be excluded. Thus, the present findings may not be generalizable beyond the hospital, including nationally or cross-culturally. Second, in order for group sizes to be sufficiently large for performing confirmatory factor analysis, we restricted our analyses to large diagnostic groups using broad ICD–10 categories. Thus, while this study included large groups of patients with diverse mental disorders, future studies are needed that include diagnostic groups that are not part of the current sample and that differentiate between specific diagnoses within the broader diagnostic categories.