Psychological well-being is markedly reduced in patients with depression. This study examined the association between self-reported sleep problems and psychological well-being in a large cohort of patients with depression prior to treatment initiation.
MethodsWe included 7051 adults with a primary diagnosis of depression or dysthymia (ICD-10 F32, F33, or F34.1) enrolled in a standardized depression treatment program in the Capital Region of Denmark between 2007 and 2022. Sleep problems were assessed using the sleep item from the Major Depression Inventory (MDI), and psychological well-being was measured by the 5-item World Health Organization Well-Being Index (WHO-5). Associations were estimated using multivariable linear regression with robust standard errors, adjusting for demographic and clinical variables including depression severity, anxiety, medication use, and somatic comorbidities.
ResultsSleep problems were highly prevalent, with only 5.5% of patients reporting no sleep problems. Complaints of too little sleep were more common than complaints of too much sleep. Self-reported sleep problems were associated with lower psychological well-being in a stepwise manner. Compared with patients without sleep problems, those reporting short sleep, long sleep or with both had lower well-being scores (β = -11.1, -9.8, and -12.5, respectively), and associations remained robust after multivariable adjustment. No interaction by sex was observed.
ConclusionIn this large clinical sample of patients with depression, self-reported sleep problems were highly prevalent and associated with graduated lower psychological well-being.
Patients with depression report markedly reduced quality of life, characterized by impairments in both emotional, social, and physical functioning.1 Poor quality of life has further been associated with an increased risk of relapse2 and suicidal ideation,3 and improving quality of life is therefore a central treatment goal in order to restore well-being and daily functioning in patients with depression.
Sleep disturbances are a core feature of depressive disorders, with up to 90% of affected individuals complaining of sleep problems.4 Traditionally, sleep problems have been regarded as a secondary symptom of depression and are included in the diagnostic criteria for depression. However, beyond their diagnostic status, sleep problems in depression have been associated with greater depression symptom severity, difficulty in treatment, and a risk factor for both incident depression and relapse.5–8 Thus, sleep problems might not only be a symptom of depression but represent an independent clinical feature that may have important consequences for the patient. Furthermore, sleep problems are potentially modifiable treatment targets and improving sleep may thus enhance treatment outcomes beyond symptom reduction.9
Previous research has also linked self-reported sleep problems to lower levels of mental well-being. Both population-based studies and studies in specific patient groups have demonstrated associations between poor sleep, reduced quality of life, and impaired mental health.10–15 Data from patients receiving specialized depression treatment remain limited, but it could suggest that sleep may contribute independently to reduced psychological well-being among individuals with depressive symptoms.
Consequently, the aim of this study was to examine the association between self-reported sleep disturbances and psychological well-being in patients with depression before treatment initiation.
Materials and methodsPopulationIn Denmark, patients with mild to moderate depression are initially treated in general practice or by private psychologists or psychiatrists. In case of more severe depression, limited effect of initial treatment or other psychiatric comorbidity, many patients are referred to outpatient hospital-based treatment in a standardized treatment program for depression. Since 2007, data on all patients referred for a psychotherapeutic treatment program in the Capital Region of Denmark have been recorded in electronic databases: the Therapy Effect Database (2007–2017) or the Treatment Effect Database (2017-onwards).16 Patients are included in the databases regardless of whether they subsequently participate in the program. Subsequent non-participation in the program may occur because the patient was deemed ineligible during the initial assessment or due to dropout before treatment start. The combined database contains data on psychiatric evaluations, treatment type and dates, and responses to standardized psychiatric questionnaires. Participants complete questionnaires prior to the initial interview, before treatment initiation (if treatment starts >10 weeks after the initial interview), and by the end of treatment.
For this study we included data on all patients in both the Therapy Effect Database and the Treatment Effect Database (hereafter collectively referred to as the Treatment Effect Database) between January 2007 and June 2022 who started a treatment program for depression. All patients without a main depression or dysthymia diagnosis (International Classification of Disease 10th edition (ICD10) codes of F32, F33 or F34.1), as well as patients below the age of 18, patients who did not start treatment, patients who did not fill out the Major Depression Inventory (MDI) before starting treatment, patients who emigrated before treatment start or foreign patients with no registration in the Danish registers were excluded. Additionally, for patients receiving multiple treatments programs, only data from the first treatment program was included. A flow chart of the inclusion and exclusion of participants can be seen in Supplementary Figure 1. We included data from the date of MDI closest to starting treatment. Supplementary Table 1 shows basic characteristics in excluded patients. However, information on excluded individuals was limited.
Data from the Treatment Effect database was linked to the following Danish nationwide registers: the Population’s Education register, the Civil Registration System,17 the Danish National Patient Registry18 and the Danish National Prescription Register.19
Sleep variables (exposure)Sleep disturbances were assessed using the Major Depression Inventory (MDI), a validated 12-item self-report questionnaire measuring depressive symptoms over the past two weeks.20 Each item is rated on a Likert scale ranging from 0 to 5, with higher scores indicating greater symptom severity.
Until 2014, the MDI included a single sleep question: “Have you had trouble sleeping at night?”. After 2014, this question was replaced by two separate questions: “Have you been sleeping too little?” and “Have you been sleeping too much?”. To ensure comparability across all assessment years, we combined these two newer questions into one measure by selecting the highest score of the two and using it as the response to the original sleep question. This approach reflects the standard interpretation of the MDI, where any sleep problem - whether too little or too much - is considered indicative of sleep problems. Using the highest score captures the most severe symptom and avoids underestimating sleep problems, which could occur if scores were averaged.
In the subset of participants who entered the treatment program after 2014 (hereafter called the “subcohort”), the revised two-question sleep item also enabled classification of the type of sleep problem. A score on each question of 0–2 was considered low, and scores of 3–5 as high. Participants were categorized as having no sleep problem if both questions were scored low, hypersomnia if “sleeping too much” was scored high and “sleeping too little” scored low, insomnia if “sleeping too little” was scored high and “sleeping too much” was scored low, and a combined sleep problem if both items were scored high (Supplementary Figure 2).
Psychological well-being (outcome)Psychological well-being was defined based on responses to the 5-item World Health Organization Well-being Index (WHO-5), which is a short rating scale consisting of 5 items measuring subjective well-being.21 Like the MDI, the WHO-5 is a self-report questionnaire of symptoms during the past two weeks and each item is treated on a Likert Scale ranging from 0 to 5. The raw score is thus 0–25, but it is usually translated to a percentage by being multiplied by 4.21 Accordingly, the WHO-5 total score in this study ranged from 0 to 100, with higher scores indicating higher psychological well-being. The mean score in the Danish general population is reported to be around 60,22 whereas the mean score in depressive individuals is typically lower than 30.22
CovariatesFrom the Treatment Effect Database, we included age, sex assigned at birth, and self-reported sick leave at baseline. From the national registers we included information on education (primary school, medium education or longer education), marital status (married, unmarried, divorced, widow/widower), depression severity (termed ICD-10 diagnostic severity) evaluated by the therapist (based on international classification of disease 10th edition (ICD10) diagnoses mild (DF32.0, 33.0), moderate (DF32.1, DF33.1) or severe depression (DF32.2–3, DF33.2–3) evaluated at first visit), previous antidepressant medication (using Anatomical Therapeutic Codes (ATC) N06A), previous sleeping medication (using ATC codes for Z-drugs (N05CF01–03), melatonin (N05CH), promethazine (R06AD02), low-dose quetiapine (N05AH04, <100 mg), or mirtazapine (N06AX11)), previous infections (ICD10 codes DA0-DB9), previous hypertension (ICD10 code DI10–11) or previous type 2 diabetes (ICD10 code DE11).
Statistical analysisLinear regression analyses were performed to calculate beta-coefficients of the association between degree of sleep problems and subjective well-being. Models were adjusted for potential confounders in a stepwise manner. To avoid overadjustment and unstable estimates from including MDI variables, we instead used ICD-10 diagnostic severity, which provide a clinically distinct measure of depression severity.
The assumptions of linear regression were systematically tested. Residual plots were inspected to assess linearity and homoscedasticity, and normal probability plots were used to examine the distribution of residuals. Multicollinearity was assessed using variance inflation factors (VIF), and influential observations were evaluated using leverage values and Cook’s distance. Tests indicated evidence of heteroscedasticity; therefore, all regression models were estimated using robust (Huber–White sandwich) standard errors to obtain valid confidence intervals and p-values. Model diagnostics were conducted for the fully adjusted model, which was the primary basis for inference. Predicted values based on the most fully adjusted model were calculated using the marginsplot command in Stata. Since previous studies have reported sex differences in the association between sleep problems and mental well-being,13 we tested for interaction by sex and interaction by ICD10 diagnostic severity by including an interaction term between the sleep item and sex or ICD10 diagnostic severity in the linear regression model. The overall significance of the interaction was assessed using a Wald test, Testparm, in Stata. We further conducted a Cuzick non-parametric trend test to formally assess the presence of a trend in psychological well-being scores across sleep disturbance categories. To examine whether sleep problems contributed independently to subjective well-being beyond overall depressive symptomatology, two nested regression models were estimated. Model 1 included the MDI subtotal score (excluding the sleep item), ICD-10 diagnostic severity, and covariates. Model 2 additionally included the MDI sleep item. The increment in explained variance and the significance of the sleep coefficient in Model 2 was used to evaluate the independent contribution of sleep problems to WHO-5.
Finally, as a supplementary descriptive analysis, we examined the association between appetite disturbance (MDI item 10 “Have you suffered from a change in appetite?”) and subjective well-being. Like sleep disturbance, appetite disturbance is a somatic/vegetative symptom and uses a similar response format as the sleep item, but it reflects a potentially different underlying mechanism.
ResultsA total of 7051 patients who started treatment in a depression program between 2007–2022 in the Capital Region of Denmark were included. Of these, 73% (n = 5164) were women and the median age was 33 years (interquartile range (IQR) 26–45 years). Depression was classified as mild (15%, n = 1020), moderate (63%, n = 4465), severe (15%, n = 1027), or unspecified, dysthymia or in remission (8%, n = 539). The median MDI score at baseline was 31 (IQR 24–36), while the median WHO5 score at baseline was 20 (IQR 12–32). In the subcohort of 5576 patients who answered both MDI items for sleep, there were a similar distribution of women (72.8%) and age (median 32.7 years). A total of 10%, 66% and 10% had a mild, moderate, or severe depression, respectively. All other variables showed a distribution similar to that observed in the full cohort. Baseline characteristics in the full cohort and in the subcohort are shown in Table 1 and mean MDI values with and without the sleep item in each sleep category are shown in Supplementary Table 2. As expected, individuals with a higher degree of sleep problems had higher mean MDI scores.
Baseline characteristics of study participants treated in a depression treatment program between 2007–2022.
*The variable does not sum up to the total number due to missing variables.
The subcohort was the subset of patients who were asked two sleep questions instead of one.
In all 7051 individuals, 17.3% reported having sleeping problems all the time, whereas 29.5% had problems most of the time, 20.1% slightly more than half the time, 11.0% slightly less than half the time, and 16.7% some of the time, while only 5.5% did not have trouble sleeping (Table 2). The mean WHO-5 score varied from 41.0 in patients without sleep problems to 16.9 in patients who reported sleep problems all the time. The predicted WHO levels after adjustment for age, sex, education or civil status, depression severity, previous anxiety or sleeping medication, sick leave at baseline, previous infections, hypertension and type 2 diabetes are shown in Fig. 1 and show that increasing sleeping problems were associated with lower psychological well-being in a stepwise manner.
Association between sleep problems and psychological well-being in 7051 patients with depression.
Model 1: Age, sex, education, civil status.
Model 2: Model 1 + depression severity, previous anxiety or sleeping medication, sick leave at baseline.
Model 3: Model 2 + previous infections, hypertension and type 2 diabetes.
MDI = Major depression inventory, WHO5 = World Health Organisation Well-Being Index, SD =Standard deviation.
Predicted levels of psychological well-being measured on the WHO-5 by sleep category in 7051 individuals with depression after adjustment for age, sex, education, marital status, depression severity, previous anxiety, previous sleeping medication, sick leave, previous infections, hypertension or type 2 diabetes mellitus. Error bars represent standard errors. Actual (unadjusted means) are shown in Table 2.
In the regression analyses, self-reported sleep problems were similarly associated with lower psychological well-being in a stepwise manner in both unadjusted and fully adjusted analyses (Table 2). Adjustment for age, sex, education or civil status did not change the beta-coefficient, while further adjustment for depression severity, previous anxiety or sleeping medication, sick leave at baseline, previous infections, hypertension and type 2 diabetes only slightly attenuated the beta-coefficients (Table 2). A significant trend was observed across sleep categories (Cuzick test, z = −31.11, p < 0.001), with WHO-5 scores decreasing progressively with increasing levels of sleep disturbance.
Stratification by sex revealed no differences in estimates between men and women (data not shown), consistent with a non-significant interaction between sex and sleep problems (Wald test, p = 0.23). For ICD10-diagnostic severity, the overall Wald test was significant (p = 0.04), suggesting that the association might vary across severity groups. The association appeared strongest in patients with mild depression and slightly weaker in patients with moderate to severe depression, but none of the severity-specific interaction terms reached statistical significance (Supplementary Table 3).
The MDI subtotal score (excluding the sleep item) was significantly associated with WHO-5 (R² = 0.483) accounting for 48% of the variance in WHO-5. When adding the MDI sleep item in Model 2, it remained a statistically significant independent predictor of WHO-5 (β = −1.24, 95% CI [−1.28; −1.20], p < 0.001), with a small increase in explained variance (R² = 0.492, ΔR² = 0.009).
Finally, the supplementary analysis of self-reported problems with appetite (MDI item 10) and psychological well-being showed a similar pattern with higher levels of appetite disturbance associated with a progressively lower WHO levels, though with lower estimates than sleep disturbance (Supplementary Table 4).
The subcohortIn the 5576 patients who responded to two sleep items introduced after 2014, the distribution of the severity of sleep problems based on the combined MDI item was similar to the severity in the full cohort, suggesting that the subcohort was representative of the overall population. In the subcohort, sleeping too little was more common than sleeping too much. Specifically, 21.2% and 14.7% reported sleeping too little most of the time or all the time, respectively, whereas the corresponding percentages for sleeping too much were 11.6% and 3.6% (Table 3).
Association between sleep problems and psychological well-being in 5576 patients with depression.
When the two questions in the updated sleep item were combined, a total of 45.5% of individuals had problems mainly with too little sleep (insomnia group with “sleeping too little” scored high and “sleeping too much” scored low), 22.1% had problems mainly with too much sleep (hypersomnia group with “sleeping too much” scored high and “sleeping too little” scored low), 3.7% had both problems (combined group with both “sleeping too much” and “sleeping too little” scored high), while 28.7% reported no or only few problems with sleeping either too much or too little (slightly less than half the time). The mean WHO score was 31.7 for patients no or light sleep problems, 20.1 for the insomnia group, 21.5 in the hypersomnia group and 18.1 in the combined group (Table 3).
In the subcohort, the severity of both types of sleeping problems, sleeping too little and sleeping too much, was associated with lower predicted WHO scores in a stepwise manner (Fig. 2), which was also seen/reflected in the regression analyses (Table 3). As in the full cohort, adjustment only attenuated effect estimates slightly. When examining the type of sleep problem, compared to patients with no or only few sleep problems, patients with any type of sleeping problem had lower predicted WHO scores. The corresponding beta-estimates were −11.1 (−12.1; −10.1) for patients in the insomnia group, −9.8 (−11.0; −8.7) for patients in the hypersomnia group and −12.5 (−14.3; −10.7) for patients in the combined group after full adjustment.
Predicted levels of psychological well-being measured on the WHO-5 by sleep category after adjustment for age, sex, education, marital status, depression severity, previous anxiety, previous sleeping medication, sick leave, previous infections, hypertension or type 2 diabetes mellitus. Error bars represent standard errors. Actual (unadjusted means) are shown in Table 2.
In this large clinical cohort of patients with depression, we had several important findings: 1) The majority of patients with depression experienced sleep problems with only 5.5% reported not having any sleep problems; 2) Sleep problems were associated with lower psychological well-being in a stepwise manner both in unadjusted and fully adjusted analyses; 3) This pattern was similar for both sleeping too little or sleeping too much, and the lower psychological well-being scores were similar in patients with symptoms of insomnia or hypersomnia; and 4) There was no interaction by sex. For depression severity, there was a tendency towards a stronger association in patients with milder depression.
The finding that sleep problems are associated with reduced well-being aligns with previous studies in the general population.13,23 In a large Korean population-based study including 225,541 adults, individuals with poor sleep reported lower quality of life compared to individual with good sleep.24 In a recent Canadian study of 30,097 community-dwelling adults, sleep problems including short sleep duration and sleep dissatisfaction were associated with life dissatisfaction, psychological distress, and poor mental health with slightly higher effect estimates in men.13 Other studies, however, have reported that the association is primarily observed in women.25 In our study, we did not find any interaction of sex. Fewer studies have specifically examined the association between sleep problems and quality of life in patients with depression. In a randomized controlled trial, McCall et al. found that comorbid insomnia in depressed outpatients was associated with markedly reduced health-related quality of life at baseline, and that targeted insomnia treatment with medication led to greater improvement in quality of life compared to antidepressant therapy alone.26 Similarly, Le et al.27 reported that the combination of insomnia and depression in a large community sample was linked to an additive reduction in quality of life, suggesting that sleep problems contribute to poorer psychological functioning among individuals with depressive symptoms. These findings align with our results, suggesting that sleep disturbance independently carry some relevance for psychological well-being beyond overall depression severity, although the general burden of depressive symptomatology remains the dominant predictor.
The tendency towards a stronger association between sleep problems and subjective well-being in patients with mild depression compared to more severe depression may not be surprising. In patients with severe depression, the overall burden of symptoms may be so severe that it dominates subjective well-being. In patients with milder depression, where the symptom burden is less overwhelming, sleep problems may represent a more prominent and relatively independent source of reduced well-being. These findings are, however, exploratory and should be interpreted with caution, as the individual interaction terms were not significant.
The association between sleep disturbances and reduced psychological well-being may be explained through several mechanisms. Psychologically, disturbed sleep reduces the capacity for emotion regulation and is associated with reduced positive affect28 and increased risk of rumination,29 which in turn exacerbates negative affect30 and thereby amplify the emotional burden of depression. Cognitive consequences of poor sleep, such as impaired attention, memory, and executive function,31 might further limit the patients’ ability to engage in work, daily activities, and social relationships, all of which are central to perceived quality of life. Biologically, sleep disruption may also contribute to circadian misalignment, alterations in stress hormone regulation, and dysregulation of the reward circuitry,32 which may aggravate anhedonia and fatigue. However, reduced well-being may also contribute to poor sleep. Evidence on the direction of this association is mixed. In a study of >6500 adolescents, bidirectional associations between sleep problems and health-related quality of life were found with similar magnitudes in both directions,33 whereas a large longitudinal study of 18,000 Finnish twins found that poor sleep consistently predicted life dissatisfaction but not vice versa.34 Thus, the direction of the association remains unclear, but a bidirectional relationship seems plausible.
A major strength of our study is the large and well-characterized sample of patients with depression, who were all included through a standardized depression “treatment program” across several years. This ensured systematic assessment procedures and data collection across participants. Because the study is based on an existing clinical database rather than voluntary recruitment, the risk of selection bias due to non-participation or self-selection is likely to be reduced. Other strengths include standardized assessment using validated scales of both depressive symptoms and symptoms of psychological well-being at baseline prior to treatment initiation, which also allowed us to evaluate both short and long sleep durations.
Limitations include the cross-sectional design which means that causality cannot be concluded. While sleep problems were associated with lower psychological well-being, the study cannot determine whether poor sleep contributes to reduced well-being, lower well-being leads to sleep disturbance, or whether both are influenced by confounding factors such as depression severity. Sleep disturbance is a core symptom of depression, and the conceptual overlap between the exposure, the outcome, and depression severity means that residual confounding likely remains despite adjustment. Although the use of ICD-10 diagnostic severity rather than the full MDI score helps avoid statistical overadjustment, it captures depression at a categorical level and may not fully account for the continuous variation in overall depressive symptomatology. The nested model analyses, showing a small independent contribution of sleep problems beyond overall depressive symptomatology, partially address this concern, but prospective designs with objective sleep measures is needed to establish the independent role of sleep disturbance. Sleep problems were assessed using only the single sleep-related item from the MDI rather than a dedicated sleep instrument. This approach does not allow differentiation between specific sleep disorders, insomnia versus hypersomnia, sleep quality, sleep duration, or objective sleep parameters, which limits the precision of our sleep measure. Furthermore, the transition after 2014 from a single general sleep item to two separate questions regarding sleeping too little or too much introduces potential heterogeneity across assessment periods. Although responses were harmonized into a single composite score, it cannot be excluded that this change in format introduced some degree of measurement inconsistency that may have attenuated or otherwise influenced the observed associations. However, the analysis in restricted to the post-2014 sub cohort yielded results consistent with those observed in the full cohort, suggesting that the impact of this measurement transition on the overall findings was limited. Furthermore, both sleep problems and psychological well-being were assessed using similar self-reported measures, which may introduce reporting bias and potentially inflate associations due to shared method variance. Although psychological well-being is subjective, the exclusive use of self-reports limits the ability to corroborate findings with objective indicators of sleep or functioning. Additionally, some patients already took antidepressant medications or sleep medication at baseline which could potentially influence the results. Furthermore, the large number of patients excluded due to not initiating treatment or lacking baseline MDI data raises the possibility of selection bias. Excluded patients were slightly more frequently male and more often had mild or unspecified/remission/dysthymia diagnoses, suggesting less severe depression overall. If patients with milder depression and fewer sleep problems were disproportionately excluded, the observed associations between sleep disturbance and psychological well-being could be somewhat overestimated. These findings should therefore be interpreted with this in mind. Finally, this study included only patients with depression and consequently may not be representative of the general population.
In conclusion, we found that self-reported sleep problems were very common in patients with depression and were associated with lower psychological well-being in a stepwise manner. While the general burden of depressive symptomatology may remain the dominant predictor of well-being, sleep disturbance demonstrated a small independent association, which along with the findings of others suggests that addressing sleep problems could be clinically relevant in the treatment of depression.
Ethical considerationsThe study was approved by the Danish Data Inspection. Data are stored Due to restrictions on the access to data stored at Statistics Denmark, raw data from the present study cannot be openly shared. Non-Danish researchers can get access to individual level data by collaboration with Danish researchers through access from computers at approved Danish research institutions.
FundingThis study was supported by the Jascha Foundation and Helsefonden, which had no involvement in the study design, data collection, analysis, or manuscript preparation.
Declaration of generative AI useArtificial intelligence tools were used to assist with language refinement and occasional coding troubleshooting. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.
All substantive content and interpretations were determined by the authors.
Author contributionsData was curated by MO and FHG. Project administration and funding acquisition was made by MWA. Formal analysis writing original draft were made by MWA and IWA. All authors were a part of conceptualization and methodology, evaluating and interpreting results, review and editing and all authors agreed to submit the paper and have approved the final version.
Marie Kim Wium-Andersen has received honoraria for lectures on sleep problems in psychiatric populations by AGB Pharma, however, these activities were entirely unrelated to the present study. The authors declare no other conflicts of interest.
