Psychoeducation for depression, anxiety and psychological distress: a meta-analysis

In this meta-analysis, a passive psychoeducational intervention is defined as an intervention which provides information, education materials or feedback/advice. Examples of passive psychoeducation are programmes offered to individuals through leaflets, posters, audio-visual aids, lectures, internet material or software which aims to educate the recipient about the nature and treatment of depressive and/or anxiety disorders or psychological distress. The intervention can be delivered in primary or secondary care settings, or within universities, community centres or other public venues. Psychoeducation can be delivered through the post, email, via face-to-face lectures or through information published on the web.

Although in some cases it is difficult to distinguish between active and passive education (that is, where encouragement is offered but no explicit instructions are given to carry out certain recommendations), in the present review, passive education was defined as education that did not require the recipient to undertake explicit homework or relaxation exercises and which did not deliver active treatment. Thus programmes which taught the principles or required the implementation of elements of active psychotherapies (for example, cognitive behavioural therapy [CBT] or interpersonal therapy [IPT]) were excluded. Studies were also excluded if psychoeducation was offered in combination with another component, such as CBT or any other broader multifactor intervention.

The Cochrane, PsycInfo and PubMed databases were searched on 25 September 2008, with the key search terms 'depress*' OR 'anxi*' OR 'psychological distress' OR 'mood' OR 'affective' OR '*phobia' OR 'OCD' or 'obsessive compulsive' OR 'panic' AND 'psychoeducation' OR 'education' OR 'information' or 'knowledge' OR 'instruction' OR 'teaching' OR 'mental health literacy' OR 'anxiety literacy' OR 'depression literacy'. In addition, the following limits for retrieving references were applied for the PubMed database: 'humans'; 'clinical trial'; 'RCT'; 'CT phase I to IV'; 'controlled clinical trial'; 'evaluation study'; and 'English'. For the PsycInfo database, the references were limited to: 'humans'; 'treatment & prevention'; 'quantitative study or treatment outcome'; 'randomized clinical trial'; and 'English'. Separate searches for systematic reviews and meta-analyses were done for the PsychINFO and Pubmed database using similar key search terms.

Two independent researchers screened the identified titles and abstracts to determine if the inclusion criteria were met. Full text copies of all potentially relevant papers which met criteria, or papers where there was insufficient information in the abstract to determine eligibility, were retrieved. Full text articles were further screened and excluded from further analysis if inclusion criteria were not met. Reference lists of all included systematic review and meta-analysis studies were checked. The data extraction of relevant papers was completed by two independent researchers, with disagreements resolved through discussion or with a third or fourth researcher.

Studies were included if: the psychoeducation targeted depression, anxiety or psychological distress; participants were described as either experiencing mood or anxiety disorders; or if they experienced elevated scores (equal to or above a specified cut-off score, see Table 1) on depression, anxiety or psychological distress scales. To be included, studies were required to have a randomized controlled design, which incorporated a no intervention, attention-placebo or a waitlist control group to which psychoeducation was compared. All included studies were required to report mental health outcomes (depression, anxiety or psychological distress) and were published in peer-reviewed, English language journals. There was no restriction on the age of participants. Studies were excluded if the education component was offered in addition to other components (for example, psychotherapy with elements of psychoeducation or psychoeducation enhanced with treatment as usual) or when the intervention was compared solely to a (potentially) active treatment (for example, medication, treatment as usual or psychotherapy). Studies were also excluded: when the intervention was not passive psychoeducation but involved an active intervention (for example, components of CBT or IPT, relaxation exercises or homework or group discussion); or when psychoeducation was aimed at target groups where there was a concomitant physical health or mental disorder; or where the target of the intervention was a carer or parent of the person with anxiety or depression (for example, medical illness, other mental health disorders, parental programmes, family-caregiver programmes).

Based on Jadad et al.'s [15] criteria, study quality was assessed against three key criteria: randomization; double-blinding; and withdrawals and dropouts. Quality ratings range from 0 to 5, although intervention trials for mental health disorders rarely are rated above 3 as double-blind conditions often cannot be achieved.

Primary outcome measures included reduction of depression, anxiety and psychological distress scores as measured on depression, anxiety or psychological distress scales. A second aim was to identify factors which may have contributed to the success of the intervention, such as the setting, the method of delivering psychoeducation, and whether the psychoeducation was based on evidence-based guidelines or research materials.

For each comparison between a psychological treatment and a control group, we calculated the effect size indicating the difference between the two groups at post-test (Cohen's d or standardized mean difference). Effect sizes were calculated by subtracting (post-test) the average score of the psychological treatment group from the average score of the comparison group, and dividing the result by the pooled standard deviations of the two groups. Effect sizes of 0.8 can be assumed to be large, while effect sizes of 0.5 are moderate and effect sizes of 0.2 are small [16].

In the calculations of effect sizes we only used those instruments that explicitly measured symptoms of depression, anxiety or psychological distress. If more than one depression measure was used, the mean of the effect sizes was calculated, so that each study only provided one effect size. If means and standard deviations were not reported, we used the procedures of the Comprehensive Meta-Analysis software (CMA [see below]) to calculate the effect size using dichotomous outcomes. Effect sizes were calculated using the differences between the psychoeducation and the control group immediately at post-test. Follow-up effect sizes could not be calculated because of the small number of data points available. In addition, the follow-up period differed considerably among these studies. Effect sizes were calculated for both completer and intention to treat data, if provided.

To calculate pooled mean effect sizes, we used the computer program CMA (version 2.2.021). As we expected considerable heterogeneity among the studies, mean effect sizes were calculated using a random effects model. In the random effects model it is assumed that the included studies are drawn from 'populations' of studies that differ from each other systematically (heterogeneity). In this model, the effect sizes resulting from included studies not only differ because of the random error within studies (as in the fixed effects model) but also because of true variation in effect size from one study to the next. For continuous variables, we used meta-regression analyses to test whether there was a significant relationship between the continuous variable and the effect size, as indicated with a Z-value and an associated P-value. As the analysed studies used different measures (both continuous and dichotomous) to indicate effectiveness, one OR was converted into Cohen's d effect sizes. The conversion from OR to Cohen's d was conducted with the method proposed by Hasselblad and Hedges [17]. It is based on the following formula: Cohen's d = v3 * LogOR/π.

The Q statistic was calculated as an indicator of homogeneity. A significant Q rejects the null hypothesis of homogeneity and shows that the variability among effect sizes is greater than what would likely have resulted from sampling error alone in the primary studies. Additionally, the I ² statistic, an indicator of heterogeneity, was calculated; 0% indicates no observed heterogeneity and larger values show increasing heterogeneity, with 25% regarded as low, 50% as moderate and 75% as high [18]. As the standardized mean difference is not easy to interpret from a clinical point of view, we transformed the standardized mean differences into the numbers needed to be treated (NNT), using the formulae provided by Kraemer and Kupfer [19]. The NNT indicates the number of patients that would need to be treated in order to generate an additional positive outcome in one of them [20]. Publication bias was tested by inspecting the funnel plot on primary outcome measures and by Duval and Tweedie's trim and fill procedure [21] which yields an estimate of the effect size after the publication bias has been taken into account (as implemented in Comprehensive Meta-analysis, version 2.2.021).

A total of 9,010 abstracts were retrieved, including 436 systematic reviews or meta-analyses. Of these, 32 papers were potentially eligible for inclusion and the full text for each of these papers was retrieved for further screening. Five papers describing four studies targeting depression and psychological distress met the review inclusion criteria [10, 22‐25]. No randomized controlled trials for anxiety were found. Nine systematic reviews or meta-analyses were further screened for possibly relevant references. On the basis of this screening, 15 full text papers were retrieved for further assessment. However, none were included for final analysis. A flow diagram is shown in Figure 1.

Of the five relevant papers, four papers describing three studies [10, 22‐24] used depressive symptoms or disorders as primary outcome measure, while one study [25] reported psychological distress as an outcome measure (see Table 1). Two studies [10, 23] used evidence-based medical/psychological depression/anxiety information; one of them also gave advice [23]. Two studies [22, 25] used mailed feedback based on test results and provided advice and one study [23] used leaflets as intervention type. Two papers reporting one study [10, 24] used a website. Two studies [10, 22] compared the intervention with an attention placebo-control, while two studies [23, 25] compared the intervention to no intervention condition. One study [10] recruited participants from the community, one study [23] used primary care participants, one study [25] recruited employees and one study [22] included college students. A total of 694 participants were recruited across all the studies. All included studies used individual rather than group formats. Interventions across all studies ranged from one single email or leaflet to six sessions of psychoeducation. See Table 1 for an overview of the included studies.

The quality of most studies was adequate. Assessors of outcomes and participants were blinded for treatment assignment in only one study. Drop-out rates varied between 4% and 17%. Two studies reported outcomes based on completer analysis [22, 25], two describing one trial reported intention to treat data (ITT) data in addition to completer data [10, 24] and one study [23] reported an odds ratio based on ITT data. Consequently, except for the converted effect size of Jacobs' [23], all effect sizes are based upon the completer's data.

The limitations of this review need to be acknowledged. First, only a small number of studies were eligible for inclusion in this review. Due to variability in population, method of recruitment, inclusion criteria, content and type of intervention, it was difficult to make comparison between studies. Therefore, factors influencing the effectiveness of an intervention were difficult to determine. Second, one study reported that a number of the participants who were included in the psychoeducation group were concurrently taking psychotropic medication. However, since some of the participants who were included in the control condition were also concurrently taking medication, it is unlikely that pharmacotherapy explains the pattern of results. Third, three reported effect sizes in this meta-analysis were based on the completer's data and one effect size was based on ITT data. Completer data is likely to yield higher effect sizes as those retained in the study may be more likely than those who dropout to show positive effects. However, this was not the case in the present study with the highest effect size associated with an ITT design. Two papers [10, 24] describing one trial provided effect sizes based on completers data as well as ITT data. This effect size (0.29) was not very different from the effect size (0.31) based on completers data. Thus, although it is not optimal to combine studies with different designs and biases, we think it unlikely that combining design types biased findings in this instance. Finally, we only included studies from peer-reviewed, English language journals. However, the effect of language bias minimally impacts the conclusions of systematic reviews [32].