Background

Prevalence and impact of depression in adults with chronic disease(s)

Depression affects 300 million people and is currently the leading cause of mortality worldwide substantially impacting social and occupational functioning [1,2,3,4]. Prevalence of depression is more common among individuals with chronic physical diseases [2, 3, 5]. Estimates indicate that approximately 20% of people with chronic physical diseases experience depression, at least twice the rate found in the general population [5,6,7].

Chronic diseases are increasingly prevalent and are currently estimated to account for 60% of all deaths worldwide [8, 9]. Studies have demonstrated that depression has a significant impact on the course and health outcomes of concomitant chronic physical diseases and complicates treatment [5, 6]. For example, depression has been shown to amplify somatic symptoms and diminish self-efficacy of health-related behaviours [6, 10, 11]. Furthermore, in addition to major depressive disorder, sub-threshold depression is associated with negative outcomes including increased morbidity in those with co-occurring physical diseases, and is also a risk factor for major depression [12,13,14].

Self-management interventions for depression

It is recognized that timely treatment of depression is likely to improve functional ability, quality of life, and will considerably reduce the burden of chronic physical disease on the care recipient [15, 16]. However, worldwide less than half of those affected by depression receive adequate treatment, as there is often limited care available or barriers to accessing it, particularly high cost and lack of mental health professionals [12, 17, 18]. In response to the need for more depression support, one cost-efficient option that has shown promise is self-management interventions. Self-management interventions for depression have been formally incorporated into many healthcare systems and are recommended in best practice guidelines [19,20,21]. The only guidelines specifically targeting the treatment of depression in those with concomitant chronic physical disease(s), recommend self-management interventions for the treatment of mild to moderate depressive symptoms or as adjunctive therapy in the case of more severe symptoms [22]. However, to our knowledge, no systematic review has evaluated the effects of depression self-management interventions across chronic physical disease populations.

Self-management refers to the tasks an individual must undertake to live well with their chronic disease(s) [23, 24]. Generally, these tasks involve: a) medical management; b) maintaining, changing, or developing new meaningful behaviours; and c) dealing with the emotional impacts of the disease(s) [25]. In self-management, the individual’s role in their own care is considered central and the individual takes on the responsibility, as much as possible, of developing the necessary skills to manage their symptoms.

Self-management includes more than providing health or disease-related information, rather it is focused on behaviour change through the development of the skills and confidence needed for successful self-management of chronic disease [26, 27]. This is usually accomplished through the use of psychoeducational or behaviour strategies [28]. Learning these self-management skills can be done independently or in collaboration with health care professionals (HCPs) or a non-professional, often peer, support person [25, 29].

Arguably the most robust conceptualization of self-management is that articulated by Drs. Lorig and Holman based on nearly 25 years of work in the area [25, 30, 31]. This theoretically informed research indicates that self-management involves core skills (e.g., problem-solving, decision-making) that can be applied across chronic diseases [25, 32, 33]. From this perspective, self-management is not considered disease specific, but is focused on the development of broad skills, and self-management programs may therefore include people with a variety of chronic diseases [32, 33]. Additional skills may also be targeted when supporting the learned management of particular diseases. Aligned with this, based on existing literature, self-management skills specific to depression have been identified including behavioural activation (e.g., increasing positive activities) and social support (see Appendix A for details and references).

Self-management, self-help, cognitive behavioural therapy, and self-care

A number of terms are often incorrectly used interchangeably with self-management, including self-help, cognitive behavioural therapy (CBT), and self-care. In comparison to self-management self-help encompasses a much broader range of interventions that are primarily self-directed (e.g., books, smartphone application). Self-help interventions are predominantly designed to limit contact with HCPs [34]. On the other hand, self-management interventions focus on specific skills and are often conducted in close collaboration with HCPs, although some may be self-directed [25]. Though many self-management interventions use principles of cognitive behavioural therapy (CBT), these are also distinct [29]. Self-management is one of the essential elements of the Chronic Care Model (CCM), an evidence-based guide to chronic disease management in primary care [26, 35]. Self-management interventions employ psychoeducational and/or behavioural strategies to assist care recipients and families develop the skills and confidence necessary to live well with a chronic disease (e.g., learning to locate resources, forming partnerships with HCPs) [25, 36]. Self-management is not a form of psychotherapy and rather than delivered by a therapist, self-management can be delivered by a variety of HCPs (e.g., it not a protected act) or as a self-directed intervention.

Finally, self-care is also frequently used interchangeably with self-management; however, they may be delineated on several fronts. Self-care involves managing one’s health, with or without a chronic condition. It also encompasses a broad range of strategies including ‘doing nothing.’ Another marked difference is that self-care does not usually involve HCP support and focuses primarily on health promotion or the prevention of disease or accident [37]. In contrast, as an element of the CCM, self-management was developed specifically for people with chronic disease and centres on the development of evidence-based skills [25, 28, 37].

Aims of this Review

Accounting for the issues outlined above, the primary objective of this review is to determine the effects of self-management interventions on reducing depressive symptomatology among adults with chronic physical disease(s) and co-occurring depressive symptoms. The secondary objectives are:

  1. a)

    To determine whether the effects of self-management interventions on the primary outcome of depressive symptoms vary depending on the participants’ co-occurring chronic physical disease and baseline level of depressive symptoms (e.g., mild, moderate).

  2. b)

    To assess whether the interventions reviewed have differential effects on the primary outcome of depressive symptoms depending on intervention characteristics and content: the type and number of self-management skills targeted, mode of delivery (e.g., face-to-face, online), type and intensity of guidance, intervention provider, format (e.g., group or individual), and duration (minutes of participation) and length of time over which the intervention was delivered.

  3. c)

    To assess the effects of self-management interventions on other physiological and psychosocial outcomes (e.g., quality of life, fatigue) among the target population.

Methods

Methodological framework

The methods for this review were developed following the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions [38] and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [39]. The protocol was registered with the prospective register of systematic reviews (PROSPERO CRD42019132215).

Criteria for considering studies for this review

Types of studies

Published or in press peer-reviewed full-text randomized controlled trials (RCTs) as well as cross-over trials and studies using a quasi-experimental design (e.g., controlled trials without randomization, but with a comparison group) were considered for inclusion in this review. To be eligible, the primary outcome (depressive symptoms) had to be measured pre- and post-intervention and only English and French language articles were included. Conference abstract and theses were excluded.

Types of participants and settings

Based on Lorig and Holman’s [25] conceptualization of self-management, participants with various chronic diseases were included. The target population for this review was adults (over age 18) with chronic physical disease(s) experiencing at least mild depressive symptoms according to a validated scale or clinical interview. The following psychometrically validated questionnaires and cut-off scores indicating at least mild depressive symptoms were eligible for inclusion:

  • Beck Depression Inventory (scores ≥10) [40, 41]

  • Beck Depression Inventory-II (scores ≥13) [42]

  • Centre for Epidemiological Studies – Depression (scores ≥16) [43, 44]

  • Centre for Epidemiological Studies – Depression 10-item version (scores ≥10) [45, 46]

  • Hamilton Depression Rating Scale – Depression subscale (scores ≥8) [47]

  • Hospital Anxiety and Depression Scale – Depression (scores ≥8) [48]

  • Patient Health Questionnaire (scores ≥5) [49]

  • Geriatric Depression Scale (scores ≥11) [50]

In terms of clinical interviews, the DSM-III (Diagnostic and Statistical Manual of Mental Disorders) and above [51,52,53,54,55] as well as the ICD-9 (International classification of diseases) and above [56, 57] were considered eligible.

Studies including participants with one or multiple chronic physical diseases as defined by the Public Health Agency of Canada (excluding dementia) and World Health Organization were included [9, 58]. Participants taking anti-depressant medication were eligible, if pharmacotherapy was not administered as part of the intervention being evaluated. In terms of co-morbidities, studies including participants with a diagnosis of bipolar disorder, post-partum depression, seasonal affective disorder, or post-traumatic stress disorder were excluded as the etiology, disease course, and treatment recommendations for these conditions differ from depression [59,60,61,62]. No limits were placed on the setting in which the study was conducted.

Types of interventions

Experimental interventions.

Studies evaluating self-management interventions for depressive symptoms were included. Interventions were eligible if they incorporated at least one of the key self-management skills described in Appendix A. These skills were derived from the theoretical literature as well as from existing self-management interventions from reputable (peer-reviewed or government issued) sources [25, 29, 63,64,65]. Further, the intervention had to be administered to the individual directly. Interventions in which the person with the chronic disease participated with someone else (e.g., family member) were excluded due to the confounding effects of social support [66].

Multi-component interventions were eligible as a long as the above self-management criteria were met and at least one component targeted mood, distress, or depression. To avoid confounding effects, studies evaluating interventions including a component in which pharmacotherapy was administered were excluded. All formats of interventions were eligible (e.g., workbooks, online modules). Level of guidance by HCPs was not an exclusion criterion; self-directed, minimal contact, and HCP administered interventions were all eligible.

Comparator interventions.

Eligible control comparisons were no treatment, treatment as usual, waitlist, and attention control groups, as long as the participants were not receiving active components of the interventions (e.g., psychological therapy or the self-management skills outlined in Appendix A).

Outcome measures

The primary outcome was depressive symptoms. The secondary outcomes were a) improvement in psychosocial outcomes such as quality of life, anxiety, self-management skills, and self-efficacy; b) improvement in physical health measures; c) improvement in alcohol and drug consumption; and/or d) decrease in health care utilization.

Information sources and study selection

Search strategies

Eligible studies were primarily identified through searches of electronic bibliographic databases. Secondary search strategies consisted of verifying the reference lists of the included full-texts and using the PubMed ‘find similar’ function. Unpublished findings were also sought through the Cochrane library as well as the national and international trial registries outlined in the Cochrane Handbook for Systematic Reviews [38]. These were searched for relevant trial protocols and, if published findings of these trials could not be found, authors were contacted directly for further information.

Database searches

Eligible studies were identified by searching the following databases: MEDLINE (1946-), EMBASE (1996 -), PsycINFO (1967 -), and Cumulative Index to Nursing & Allied Health (CINAHL) (1984 -). The search was conducted in June 2018 and no limits were applied. All databases were searched using a combination of keywords and subject headings across three concepts: a) self-management, b) depression, and c) trial design (RCT or quasi-experimental). The search strategy was developed in consultation with a health sciences librarian and were assessed using the Peer Review for Electronic Search Strategies (PRESS) guidelines [67]. The full primary electronic search for Ovid Medline is included in Appendix B. All titles and abstracts were downloaded to a citation manager, EndNote, and screened using Rayyan online software. Duplicates were removed according to the procedures outlined in Bramer, Giustini [68].

Study selection

Two authors independently assessed the eligibility of all retrieved titles, abstracts, and full texts to confirm inclusion or exclusion. Two authors also separately examined the reference lists of included full texts. Finally, searches were conducted using the ‘find similar’ function of PubMed. Any disagreements were discussed with a third author until consensus was established. Multiple reports relating to the same study were aggregated so that findings reflected each study rather than each report.

Data extraction

Data were extracted using a standardized Microsoft Excel form that was developed based on the Cochrane Handbook for Systematic Reviews [38]. The form was adapted from one used in a previous systematic review conducted by team members [69, 70]. Data were extracted by one author and confirmed by at least one other. Disagreements were discussed with a third author until resolved.

The following data on study characteristics were extracted: citation details, country of origin, study design, aims, theoretical framework, population (age, diagnosis, gender, depressive symptoms), sample size, setting (e.g., hospital, community), summary of intervention and control groups, self-management skills in the intervention, format of the intervention, intervention provider, level of guidance, duration of the intervention (number of minutes of participation), length of intervention (time period over which the intervention was delivered, e.g., 2-months of sessions), monitoring of fidelity and adherence, outcomes, timing of measurement, and attrition [38]. Outcomes were grouped into two time periods: T1 from baseline < 6-months post-baseline and T2 ≥ 6-months post-baseline. If any data were missing or unclear, the authors of the manuscript were contacted for further information.

Methodological quality

Two authors (combination of the first, seventh, and eight authors) independently assessed the risk of bias of each study based on the criteria outlined in the Cochrane Risk of Bias tool [71]. Again, disagreements were discussed with a third author until consensus was reached. The risk of bias was evaluated according to the following criteria: a) inclusion criteria specified, b) pre-specified primary outcome(s), c) psychometric properties of primary outcomes provided, d) explicit power calculation, e) target sample size reached, f) appropriate randomization procedures and allocation concealment, g) discussion of potential co-interventions, h) baseline characteristics of all groups provided, i) blinding of outcome assessors, participants, and interventionists, j) adherence to intervention (> 75%), k) fidelity monitoring, l) management of missing data (intention-to-treat analysis), m) participant retention (> 80%), and n) reasons for attrition stated.

Each potential source of bias was evaluated as having been met (score 1) or not met (score 0). If the information was not specified in the manuscript the authors were contacted. If the information could not be clarified, the item was deemed not to have been met. Studies were considered to be of high methodological quality if 13–17 of the criteria were met, moderate quality if 8–12 were met, and low fewer than 8 were met. Direct quotes from each study as well as supporting comments were included in the evaluation of each. For evidence of selective reporting, study protocols or trial registration of included articles, when available, were compared to the published findings and unexplained discrepancies were noted [71].

Data analysis

Effect sizes (Hedge’s adjusted g) were calculated using outcome scores at post-intervention assessment between treatment and control conditions for the T1 and T2 time periods [38, 72]. Initially analyses were planned for three time periods: T1< 3 months post-baseline, T2 - 3 to < 6 months post-baseline, and T3 ≥ 6 months post-baseline. Due to the limited sample size, T1 and T2 were combined. Hedge’s adjusted g was selected to reduce potential bias due to small sample sizes [73]. The magnitude of the effect size can be interpreted according to the benchmarks outlined by Cohen (1988) [74]; namely, small (0.2), moderate (0.5), and large (0.8).

Pooled mean effect sizes were calculated to obtain a summary statistic for the T1 and T2 periods. If a study reported both per protocol and intention-to-treat analyses, per protocol data were included in the meta-analysis. When outcome data were collected at more than one time point within the timeframe, as most studies reported data at 6-months, the data collection closest to this point was used (e.g., if 6 and 12 month post-baseline data were available, 6 month data were entered into the meta-analysis). The Higgin’s statistic (I2) was calculated to measure the heterogeneity across studies and interpreted as 0% indicating no heterogeneity, 25% low, 50% moderate, and > 75% high heterogeneity [71]. It was anticipated that there would be considerable heterogeneity across studies arising from differences in the interventions delivered, sample characteristics, and study designs. As such, a random effects model was used for meta-analysis calculations. As the number of studies was small, in addition to the DerSimonian–Laird approach, the Knapp-Hartung approach was used to make small-sample adjustments to the variance estimates of any outcomes that were statistically significant [75]. This more conservative approach produces a wider confidence limit appropriate when the sample size is small [75]. All tests were two-sided, and the significance level was set at p < 0.05. These analyses were conducted using RevMan 5.3.

Publication bias was evaluated through inspection of funnel plots of the primary outcome variable of depression [38]. The effect of potential moderators on the primary outcomes was also assessed [76]. The prespecified moderators, participant and intervention characteristics, are detailed in Table 1. Studies were separated into sub-groups based on these variables and meta-regressions were performed to identify whether there was a statistically significant difference in outcomes between the sub-groups. There is no consensus on the number of studies required to run a meta-regression; these were performed when sub-groups included 4 or more studies (p-value < 0.05 was set to establish significance) [77]. As meta-regressions cannot be performed using the Revman 5.3 software, these were calculated using the ‘metareg’ program in STATA (version 15.1). If needed information for any of the above calculations was not reported in the publication, authors were contacted for further details. If data required for inclusion in the meta-analysis were not available, the study was included for descriptive review only. Due to the substantial diversity in reported outcomes, only outcomes reported at least three times at one time point (T1 or T2) were included for review.

Table 1 Descriptive summary of included studies
Full size table

Results

Study selection

In total, 21,663 titles were retrieved through database searches and over 500 titles were screened through secondary searches. After removing duplicates, 19,788 titles remained. Screening of these resulted in the inclusion of 2212 abstracts, 1832 of which were excluded, leaving 380 full texts to be reviewed. Seventeen manuscripts reporting on 15 studies were retained: 12 for inclusion in the meta-analysis and three for descriptive review only. Flow of studies and reasons for exclusion are detailed in Fig. 1.

Fig. 1
figure 1

PRISMA 2009 Flow Diagram

Full size image

Description of studies

Characteristics of included studies are described in Table 2 and there was no indication of publication bias (Appendix C). Studies were conducted in the United Kingdom (n = 5), United States (n = 3), the Netherlands (n = 2), Germany (n = 2), Iran (n = 1), Republic of Korea (n = 1), and Australia (n = 1). Eleven studies used a 2-group RCT design (including two pilot trials), two studies used a 2-group quasi-experimental design, and one used a 3-group RCT design. The intervention groups were compared to usual care (n = 6), attention control groups (n = 6) (e.g., provided publicly available health information), or waitlist control (n = 3).

Table 2 Effect sizes for T1 and T2 for secondary outcomes
Full size table

Participants

In total, 2064 participants were included in this review, with study sample sizes ranging from 40 [91] to 500 [94]. Most studies (n = 12) included more women than men and two studies included only women [90, 96]. Participants’ primary diagnoses were cancer (n = 7) [80, 89, 93,94,95,96,97], diabetes type II (n = 4) [84, 88, 90, 91], chronic heart disease (n = 1) [78], chronic obstructive pulmonary disorder (n = 1) [84], multiple sclerosis (MS) (n = 1) [83], epilepsy (n = 1) [92], and chronic kidney disease (n = 1) [82] (non-exclusive categories as some studies focused on two disease groups). Mean reported age in the sample ranged from 35.0 to 70.8 years. Depressive symptomatology across study groups ranged from mild to severe, with the mean reported symptoms most often in the moderate range [41,42,43,44, 47,48,49, 98, 99].

Interventions

Included interventions are described in Table 2. Six of the 15 studies evaluated the same or similar interventions in different populations (4 one type of intervention and 2 another) [83, 92,93,94,95, 97]. Time spent participating in interventions (duration) ranged from 20 [88] to 2340 min [96] (n = 14, mean = 552.9, SD = 662.9). The length of time over which interventions were delivered ranged from one session [88] to two 12 month programs [89, 96].

The primary format of the interventions was individual (n = 11) [78, 80, 83, 84, 88, 91,92,93,94,95, 97]; however, three studies used a group format [82, 90, 96], and one included both group and individual sessions [89]. In terms of mode of delivery, seven studies used a combination of face-to-face and telephone contact (most favoured face-to-face contact with telephone follow-up only if needed) [78, 91, 93,94,95,96,97], four were delivered entirely face-to-face [82, 84, 89, 90], three were online [80, 83, 92], and one was a video on a computer tablet [88].

In terms of level of guidance, most were led by an interventionist (n = 11) [78, 82, 84, 89,90,91, 93,94,95,96,97], three were self-directed (two online programs and one video) [83, 88, 92], and one intervention was guided self-directed (participants independently worked through the intervention with feedback on exercises) [80]. The interventionists were all HCPs, other than one provided by a trained research assistant supervised by a psychologist [91]. Four of the interventions were delivered by nurses supervised or supported by psychiatrists [93,94,95, 97], three were delivered solely by nurses [78, 84, 90], one by psychologists [96], and the remaining two interventions were delivered by interdisciplinary HCPs [80, 82].

The content of the interventions focused on a combination of structured problem-solving (n = 12) [78, 80, 84, 89, 91,92,93,94,95,96,97], providing disease specific health information (n = 13; it was an optional component in two of the 13 interventions) [78, 82, 88,89,90,91,92,93,94,95,96,97], relaxation and stress management (n = 7) [82, 83, 88,89,90, 92, 96], using CBT principles (e.g., challenging negative self-talk) (n = 5) [83, 84, 89, 90, 92], care coordination (e.g., communicating the depressive symptoms to HCP team members) (n = 4) [93,94,95, 97], and finding health services (n = 2) [78, 88].

Eleven interventions were coded for a possible 13 depression self-management skills (interventions that contained the same content were combined). The skills identified for each intervention are summarized in Appendix A. Across the sample, the mean number of skills was 6.2 (SD = 2.4, range 2–11). The most frequently included skill was problem-solving (n = 9), followed by decision-making (n = 8), taking action (n = 7), social support (n = 7), and self-tailoring (n = 7). Less frequently addressed skills were social support (n = 1), resource utilization (n = 2), and forming partnerships with HCPs (n = 2).

Methodological quality

Quality assessment scores are included in Table 2 and detailed scoring is available in Appendix D. The mean quality assessment score across the sample was 11.2 (SD 2.14) out of a possible 17 indicating that on average the studies were of moderate methodological quality. Scores ranged from 6 to 14 with four studies assessed as being of high methodological quality [91, 94, 95, 97]. The least met criteria related to blinding.

Outcomes: Descriptive and meta-analysis

Primary outcome: Depression

Eleven studies were included in the meta-analysis for the T1 period (see Fig. 2). The pooled effect size of − 0.47 [95% CI -0.73, − 0.21] was significant with high heterogeneity I2 = 76% and favoured the interventions over the control conditions. The results remained significant after the Knapp-Hartung (KH) conversion [95% CI -0.74, − 0.02]. Statistically significant effect sizes ranged from − 0.78 [90] to − 1.13 [88].

Fig. 2
figure 2

Forest Plot of Depression T1 - Baseline to < 6 months

Full size image

When examining potential sources of heterogeneity, three studies reported pharmacological co-intervention that was significantly imbalanced between the intervention and control groups (participants in the intervention group were more likely to begin pharmacotherapy than those in the control group) [78, 93, 94]. When these studies were removed from the meta-analysis, the pooled effect size with 8 studies was of − 0.41 [95% CI − 0.61, − 0.20] with I2 = 32%. Using both the DerSimonian–Laird and the more conservative KH approach and examining potential sources of heterogeneity, the findings of all analyses indicated a statistically significant moderate effect of interventions as compared to control conditions. The two studies not included in the meta-analysis measuring depression outcomes reporting findings in line with those of the meta-analysis [95, 96].

In the T2 time period, 7 studies were included in the meta-analysis (see Fig. 3). The pooled effect size of − 0.53 [95% CI -0.91, − 0.15] was statistically significant with high heterogeneity, I2 = 86% favouring the interventions. The results remained significant after KH conversion [95% CI -95, − 0.13]. Excluding the same three studies as in T1 [78, 93, 94] from the meta-analysis resulted in a pooled effect size, with 4 studies, of − 0.53 [95% CI -0.84, − 0.21] with moderate heterogeneity, I2 = 50%. The one study [97] not entered into the meta-analysis at T2 also favored the intervention over control group.

Fig. 3
figure 3

Forest Plot Depression T2 ≥ 6 months post-baseline

Full size image

Secondary outcomes

A summary of results for secondary outcomes is presented in Table 3. Forest plots of meta-analysis results at the T1 and T2 periods for secondary outcomes are in Appendix E.

Table 3 Moderator analyses outcomes T1
Full size table

Anxiety.

In the T1 period, 7 studies were entered in the meta-analysis [78, 82, 86, 88, 90, 93, 94]. The pooled effect size was − 0.42 [95% CI -0.73, − 0.12] with heterogeneity of I2 = 73% in favour of the interventions. This finding remained significant after HK conversion [95% CI -0.82, − 0.02]. Removing the same three studies as identified for depression due to high pharmacological co-intervention [78, 93, 94], the pooled effect size was significant, − 0.29 [95% CI -0.53, − 0.06], with no heterogeneity, I2 = 0%. One study [95] was not entered into the meta-analysis and favoured the intervention.

For the T2 time period, 4 studies were included in the meta-analysis [78, 90, 93, 94]. The pooled effect size was − 0.52 [95% CI -0.94, − 0.10] with high heterogeneity, I2 = 77%. This was not significant after HK conversion. The one study [97] not entered into the meta-analysis reported a significant improvement in anxiety in the intervention group as compared to the control.

Health-related quality of life.

Mental component score (MCS) of health-related quality of life. Five studies were included in the meta-analysis in the T1 period [78, 80, 86, 88, 90]. The pooled effect size was statistically significant with moderate heterogeneity, 0.43 [0.09, 0.76] with I2 = 60%. However, it was not significant after HK conversion. Removing the results reported by Barley et al., 2014 [78] did not improve heterogeneity. For T2, the 3 studies were entered in the meta-analysis and did not result in significant pooled effect sizes [84, 90, 91].

Physical component score. (PCS) of health-related quality of life Five studies were entered into the meta-analysis for the T1 period [78, 80, 84, 88, 90] and 3 studies for the T2 time period [78, 84, 90]. The pooled effect sizes were not significant at either time point.

Fatigue.

For the T1 period, three studies were included in the meta-analysis [80, 83, 94] and resulted in a significant pooled effect size of − 0.36 [− 0.67, − 0.06] with moderate heterogeneity, I2 = 50%. The results were not significant if Sharpe, Walker [94] was removed from the analysis. For T2, only one study reported the needed data for meta-analysis [94] and the results were statistically significant with an effect size of − 0.60 [− 0.78, − 0.41] in favour of the intervention. Of the three studies not included in the meta-analysis, two reported in favour of the intervention at the T1 and T2 time periods [95, 96], and the remaining study reported no significant effect on this outcome [97].

Glycemic control (HbA1c).

Three studies were entered into the meta-analysis in the T1 period [87, 90, 91] and the results of the pooled effect size were not significant. At T2, 4 studies were included in the meta-analysis [87, 89,90,91] and the results were significant with an effect size of − 0.35 [CI 95% -0.62, − 0.07] and no heterogeneity, I2 = 0%.

Moderator analyses

The results of the moderator analyses for the T1 time period are presented in Table 3. The three studies found to be outliers [78, 93, 94] due to imbalanced pharmacological co-intervention were not included in these analyses. There was not enough data to perform meta-regression for the T2 period (4 studies total); however, 8 studies were included at T1. Meta-regressions were performed for the following 5 moderators, as there were 4 studies in each sub-group: duration of the intervention (< 300 min or ≥ 300 min), behavioural activation (yes/no), health habits (yes/no), self-tailoring (yes/no), and number of self-management skills included in the intervention [1,2,3,4,5,6 or 7,8,9,10,11,12,13]. None were found to be significant. Additional meta-regressions were run for the following 9 moderators including a minimum of 3 studies per sub-group: baseline depression level of the study sample (mild to moderate/moderately severe to severe), level of guidance (guided/self-directed), intervention provider (professional/self-directed), length of the intervention (< 3 months/ ≥ 3 months), decision-making (yes/no), taking action (yes/no), cognitive restructuring (yes/no), self-monitoring (yes/no), and relaxation (yes/no). Of these, the results were significant only for the two self-management skills of decision-making (p = 0.020) and taking action (p = 0.017).

Discussion

To our knowledge, this is the first systematic review to examine the effect of self-management interventions on reducing depressive symptoms among adults with chronic physical disease(s) and co-occurring depression. The results were drawn from the findings of 15 studies. Meta-analysis was conducted for two time periods for the primary outcome of depression as well as the secondary outcomes of anxiety, health-related quality of life (mental component and physical component), fatigue, and glycemic control. Analyses of potential moderators of intervention effect on the primary outcome were performed to identify active elements. Overall, the findings support: a) an effect of interventions on improving depression and anxiety as well as glycemic control at ≥ 6-months post-baseline; b) intervention duration and intervention length were not found to impact effect; and c) some moderators merit further attention in future studies, including level of guidance and type of self-management skills included.

Effect on participant outcomes

The results of the review indicate a moderate effect of self-management interventions on depression and a small effect on anxiety. These results are consistent with the broader literature on self-management for individuals with depression or chronic physical diseases. A systematic review by Houle et al. (2013) found that self-management interventions reduced depressive symptomatology in the general adult population and improved functioning, self-efficacy, and self-management behaviours [65]. Findings related to relapse of depression were mixed. The results of the present review could not shed further light on this issue, as only three studies reported intervention and control group data at 12 or more months [78, 89, 93]. Another recent systematic review assessing the effect of face-to-face self-management interventions for adults with a chronic diseases (not necessarily with co-morbid depression) found that efficacious interventions were more likely to include psychological coping or stress management strategies [27].

Results of the present review can also be compared to those of a review of psychotherapy for adults with depression (with or without chronic co-morbid chronic conditions) that found the interventions to have a larger but still moderate effect size (d = 0.68) in improving depressive symptoms [100]. Reviews of the effect of psychotherapy on co-morbid depression in adults with chronic physical diseases report between small and large effect sizes [101]. Due to the relatively small number of studies, strong comparisons or conclusions cannot be drawn; however, results suggest that depression self-management interventions for this population may have a similarly beneficial effect while generally being more cost-effective.

Moderator analyses

The analyses of moderators found no significant difference in intervention effect on depression based on intervention length or duration. These findings have important implications for the integration of such interventions into resource constrained clinical environments. Seeking evidence-based cost- and time-effective interventions is imperative for the sustainability of providing these in clinical practice [27]. As feasibility is a priority, further investigation of these intervention characteristics is warranted.

Self-directed interventions (no contact with an HCP or coach) did not have significantly lower effect sizes than guided ones. However, this must be interpreted with much caution as the sample size was very small. A number of reviews support the efficacy of minimally guided and self-directed interventions [102,103,104,105, 120]. A review of self-directed psychological interventions found a small significant effect of interventions (d = 0.23 at post-test, d = 0.28 from 4- to 12-months) on depression [121]. Only one trial to our knowledge has directly compared the effects of a guided (coached) and self-directed depression intervention for adults with chronic conditions [106]. This trial was excluded from our meta-analysis because of the active control group. The results indicated an overall significant benefit of coaching on depressive symptoms at 3 months; among those who were not receiving psychological treatment at study entry, the benefit was extended to 6 months. The increased effect is likely explained by the greater adherence to the self-care tools in the coached (guided) group [22, 103].

The results indicated that not all self-management skills may be equally beneficial in improving depressive symptoms. Of the 13 self-management skills examined, two of them, decision-making and taking action, were potentially significant moderators of the primary outcome of depression. Of the skills examined, six, drawn from the work of Lorig & Holman (2003), are considered “core” self-management skills that are applicable across chronic diseases. The remaining seven skills were drawn from the literature specifically on the self-management of depression. The findings of this study parallel those of a previous review of non-pharmacological depression interventions for caregivers, which found that problem-solving, decision-making, and taking action were significant moderators of depression [70]. It is notable that neither review identified depression specific skills as significant moderators. Interestingly, a review of self-management interventions in a different population, adults with low income or low health literacy, also found that problem-solving and taking action were more often included in efficacious interventions [69]. Due to limited data, it was not possible to examine problem-solving in this review. Together, the findings of these reviews suggest that developing core self-management skills to foster behaviour change might be more important than disease-specific self-management skills.

Methodological quality

The majority of studies were of moderate methodological quality, and none met all of the criteria. Due to insufficient sample size, it was not possible to examine the heterogeneity among studies of different methodological quality. Adherence rates were reported in nine of 15 studies, similar to a previous review of self-care interventions for anxiety or depression that found 55% of studies reported adherence measures [107]. A number of studies described the amount of time participants engaged in the intervention; however, no study indicated a minimal therapeutic dose or exposure to the intervention. More detailed standardized measures of adherence including activity or module completion, time spent, and active engagement have been proposed to address this issue [108]. The pre-established criteria, based on Cochrane Risk of Bias tool, were difficult to apply to self-directed interventions. For example, intervention fidelity becomes essentially synonymous with adherence in the case of self-directed interventions. In this case, blinding of participants to group allocation may be of greater importance as they are in essence the interventionists and frequently self-report their own outcomes.

If attrition was 20% or less across the sample, the criterion was assessed as being met [71, 109]. However, in six of the 15 studies, attrition rates were notably higher in the intervention group as compared to the control [78, 84, 89,90,91,92] and no studies reported greater attrition in the control group. This raises concern regarding the potential impact of attritional bias on study outcomes [110]. Though four of these studies used intention-to-treat analyses, this may not entirely mitigate the impacts of this missing data [111]. Reporting the baseline characteristics of participants who were and were not included in analyses is recommended to help address this [111]. Further, previous work has addressed predictors of attrition of participants with depression from pharmacological trials, but this has not been thoroughly addressed in psychosocial trials [112, 113].

Reported outcomes

There was substantial variety in the outcomes reported across studies and measurement instruments used. Of the 38 outcomes measured across studies, only five were reported at least three times at one time point. Further, six instruments were used to measure the primary outcome of depression. These instruments were also used to establish the presence of at least mild symptoms of depression across the sample, a criterion for inclusion in this review. This is notable as a recent co-calibration study examining the variations in five commonly used depression self-report instruments found that the cut-off scores across scales were not equivalent [114]. The impact of this on the primary outcome could have resulted in an under- or over-estimation of intervention effect.

Strengths and limitations

The study methods were guided by the Cochrane Handbook and the PRISMA statement, and the protocol was registered with PROSPERO [38, 39]. The methods were outlined in detail and are reproducible. Given some terminological ambiguity in the literature regarding what constitutes a self-management intervention, another strength of this review was the search terms applied were very inclusive. Interventions that were not self-described as self-management were included in the review based on the meeting the predetermined definition based on current self-management literature. This is in line with recommendations by Lorig and Holman (2003). However, due to the variations in definitions of self-management, it is possible that other teams would have identified different interventions for inclusion. Further, self-management interventions are recommended as part of a stepped care approach to depression management in which a sequence of treatments are provided based on an individual’s response [115]. Lower intensity interventions, such as self-management, are initially delivered and treatment is ‘stepped up’ to higher intensity psychological interventions (e.g., pharmacotherapy, psychotherapy) for those who do not benefit from the initial treatment [22, 105, 115, 116]. Though four such interventions were screened at the full text stage, none were included as they did not meet the a priori inclusion criterion of providing non-therapeutically active control group data. Future research could evaluate the role of self-management within the structure of stepped care approach to intervention delivery.

The results should be interpreted with caution as the sample size was small with substantial heterogeneity, though this was found to be largely attributable to the confounding impact of pharmacotherapy. The limited number of studies prevented further examination of sources of heterogeneity and analyses of moderators was also conducted with a very small sample. There was only sufficient data to examine outcomes up to 6 months; longer-term outcome data is needed. Further, most of the included studies were focused on those with cancer or diabetes. Though the findings offer potential future avenues for exploration, further evidence is required to investigate longer-term outcomes, sources of heterogeneity, and possible differences in chronic physical disease populations.

Conclusion

This is the first systematic review to examine the effect of self-management interventions on depression in adults with co-occurring chronic physical diseases. The findings indicate the interventions reduced depression with a moderate effect size and anxiety with a small effect size. Impact of the interventions on other psychosocial and physical health outcomes was mixed. Recommendations include further evaluation of the impact on the amount of guidance, length, and duration of interventions, as self-directed, shorter and thereby less resource intensive interventions may be effective. Including self-management skills of decision-making and taking action in future interventions is also recommended.