Effectiveness of psychoeducation for relapse, symptoms, knowledge, adherence and functioning in psychotic disorders: A meta-analysis

Schizophrenia Research 96 (2007) 232 – 245 www.elsevier.com/locate/schres

Effectiveness of psychoeducation for relapse, symptoms, knowledge, adherence and functioning in psychotic disorders: A meta-analysis T.M. Lincoln ⁎, K. Wilhelm, Y. Nestoriuc Faculty of Psychology, Philipps-Universität Marburg, Germany Received 6 May 2007; received in revised form 21 July 2007; accepted 23 July 2007 Available online 7 September 2007

Abstract Psychoeducation (PE) for schizophrenia and other psychotic disorders is widely adopted but insufficiently evaluated. So far, meta-analytic data has demonstrated efficacy for PE when interventions include family members. Whether PE directed solely at patients is also effective remains unclear. The current meta-analysis evaluates short- and long-term efficacy of PE with and without inclusion of families with regard to relapse, symptom-reduction, knowledge, medication adherence, and functioning. Randomized controlled trials comparing PE to standard care or non-specific interventions were included. A literature search in the Cochrane Library, PsycINFO and Medline retrieved 199 studies for closer examination, of which 18 studies, reporting on 19 comparisons, met the inclusion criteria. These studies were coded with regard to methodology, participants, interventions and validity. Effect sizes were integrated using the fixed effects model for homogeneous effects and the random effects model for heterogeneous effects. Independent of treatment modality, PE produced a medium effect at post-treatment for relapse and a small effect size for knowledge. PE had no effect on symptoms, functioning and medication adherence. Effect sizes for relapse and rehospitalization remained significant for 12 months after treatment but failed significance for longer follow-up periods. Interventions that included families were more effective in reducing symptoms by the end of treatment and preventing relapse at 7–12 month follow-up. Effects achieved for PE directed at patients alone were not significant. It is concluded that the additional effort of integrating families in PE is worthwhile, while patient-focused interventions alone need further improvement and research. © 2007 Elsevier B.V. All rights reserved. Keywords: Psychoeducation; Schizophrenia; Psychotic disorders; Effectiveness

⁎ Corresponding author. Section for Clinical Psychology and Psychotherapy, Faculty of Psychology, Philipps-Universität Marburg, Gutenbergstr. 18, 35032 Marburg, Germany. Tel.: +49 6421 2823647; fax: +49 6421 2828904. E-mail address: [email protected] (T.M. Lincoln). 0920-9964/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.schres.2007.07.022

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

1. Background Psychoeducation (PE) is a frequently applied intervention for psychotic disorders in German speaking European countries. PE has been defined as systematic didactic-psychotherapeutic intervention, designed to inform patients and their relatives about the disorder and promote coping (Wiedemann et al., 2003). In a recent survey of all psychiatric institutions in Germany, Austria and Switzerland 72% of the responding institutions reported that they had offered PE for schizophrenia in the year 2003, mostly directed at patients during inpatient stay (Rummel-Kluge et al., 2006). 1.1. State of research on PE, limitations and open questions Several meta-analyses have demonstrated the efficacy of family interventions which include PE as one component among others, such as communication, social skill training or problem solving skill training (Barbato and D`Avanzo, 2000; Mari and Streiner, 1994; Pfammatter et al., 2006; Pharoa et al., 2006; Pilling et al., 2002a,b; Wunderlich et al., 1996). However, the combination of PE with additional interventions makes it hard to specify its contribution to outcome and failure to evaluate treatment components on their own increases the risk of using sparse funding on a mixture of efficacious and useless treatment ingredients (Merinder, 2000). Moreover, some of the additional components included in PE-packages, such as social skill training (Pfammatter et al., 2006) or relapse prevention (Mueser et al., 2002) have been shown to be effective on their own, while the effectiveness of the educational element of conveying knowledge on the etiology and symptoms of the disorder is less clear. Two meta-analyses focussed more directly on PE: A cochrane review (Pekkala and Merinder, 2004) analyzed 10 randomized controlled studies on PE compared to standard care, of which nine included the patients' relatives. PE significantly reduced relapse or readmission rates at nine to 18 month follow-up. The results also indicated that PE could have a positive impact on knowledge gain, adherence to medication and global level of functioning, but these analyses were based on very small numbers of studies using measures that the authors found difficult to interpret. Pitschel-Walz et al. (2001) meta-analytically evaluated the effects of 25 studies on the effect of PE family interventions in reducing relapse. They found a small significant effect for the interventions compared to standard care. However, even in these analyses the authors did not specify quality

233

or quantity of PE according to the definition above in their inclusion criteria. As a consequence, Pekkala and Merinder included interventions that consisted purely of medication counseling (e.g. Goulet et al., 1993; Razali and Yahya, 1995) while Pitschel-Walz included a row of studies in which PE was only a minor part of the intervention (Hogarty et al., 1997; Randolph et al., 1994; Vaughan et al., 1992). Mueser and colleagues (2002) reviewed four randomized controlled trials on psychoeducation-programs and found three of them to increase participants' knowledge but only one to have a positive effect on medication adherence. Another limitation of the present state of research is that meta-analyses and reviews have generally focused on family PE rather than patient directed PE or have not separated studies by focus of intervention. The two exceptions to this have yielded inconclusive results: Pitschel-Walz et al. (2001) analyzed six studies that compared family intervention with patient interventions. They found the integration of effect sizes to reveal no significant differences at post-assessment and the results for follow up assessments to be inconclusive. The authors point out that conclusions are difficult to draw due to the low number and the heterogeneity of studies. Merinder (2000) reviewed seven randomized controlled studies on the efficacy of patient education. While most of these studies reported a significant effect of PE on knowledge, positive effects were scarce for compliance, relapse and symptoms, but here too interpretation was complicated by methodological limitations and insufficient reporting. Thus, there is still no meta-analytically proven efficacy for PE that is directed at patients only. This does not correspond well with clinical practice in which the more readily conducted patient-directed approaches clearly dominate. For example, of the responding institutions in the investigation carried out by Rummel-Kluge et al. (2006) 33% offered PE for patients and relatives, 1% offered PE for relatives only and 66% offered PE for patients only. In order to provide clear guidelines for clinical practice it is important to establish more definitely whether PE offered to patients alone is an effective treatment compared to standard care and how much can be gained by the additional effort of involving the family. Finally, most studies evaluating PE have focused on knowledge and adherence, symptom reduction and relapse prevention. Hence, these outcome variables are reflected in prior meta-analyses. Increasing patients' knowledge about the disorder and its treatment can be expected to lead to better collaboration with professionals, promote adherence and self-management of

234

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

(2) Investigated patients were required to have the following diagnoses: schizophrenia, schizoaffective disorder, delusional disorder, short psychotic disorder or schizotype disorder according to a diagnostic classification scheme. (3) The comparison group was required to be either non-active such as waiting-list (WL) or treatmentas-usual (TAU), or a non-specific intervention (NSI) without proven effectiveness (e.g. problemsolving; supportive treatment, leisure time groups). Studies with active control-groups (e.g. CBT or other forms of PE) were excluded. (4) Studies were required to describe a treatment protocol in which PE was the core element of treatment (conducted in more than 50% of the total treatment time). PE was defined according to the definition by the Arbeitsgruppe Psychoedukation (Wiedemann et al., 2003), with a focus on conveying relevant information about the disorder and its treatment while promoting better coping. Specifically, conveying of information had to be a major component which could, but need not be supplemented by behavioral interventions (e.g. enhancing coping strategies and early symptom monitoring) and had to take place in a direct therapist–patient interaction. Studies using solely internet programs or written material were excluded. (5) Treatment outcome had to be reported for relapse or rehospitalization, symptoms, functioning, knowledge about the disorder or treatment adherence. (6) Studies were required to provide statistical data that allow to estimate effect sizes: means and standard deviations, t- or F-values, change scores, frequencies or probability levels. (7) Studies had to report in English, German or French.

symptoms and thus reduce symptom severity and susceptibility to relapses (Mueser et al., 2002). While these outcomes thus remain important, there is growing agreement that relevant changes in clinical performance should also be having an impact on functional outcome (Mari and Streiner, 1994) as such improvement allows patients to spend less time with symptoms and more time pursuing their goals (Mueser et al., 2002). The present meta-analysis aims at evaluating the efficacy of interventions for schizophrenia and other psychotic disorders in which PE is the primary element, with and without integration of family members, on knowledge about the disorder, adherence, relapse and rehospitalization, symptoms and functioning. 2. Method 2.1. Search procedure Relevant studies were identified by an electronic literature search in the databases PsycINFO, Cochrane Library and Medline for articles published from the first available year to March 2006. We used the search terms psychoeducation⁎, or (patient/family)education or education⁎ combined with either schizophrenia, schizophren⁎, or psychosis in the complete text of the article. This procedure was adopted to ensure that interventions with broader labels than psychoeducation (e.g. illness management or psychosocial rehabilitation) would not be missed if they included educational elements. Additional studies were identified by a manual search in Schizophrenia Research and Schizophrenia Bulletin from 1995 to March 2006 and in reference lists of reviews and meta-analyses on PE for psychotic disorders. A priori decision was made to search only for published work and to control for publication bias via posteriori analysis. Together, these searches generated 2952 studies. Of these, 2752 publications could be excluded after screening the title. In these cases the title made clear without doubt that the article was not relevant to the research question, e.g. by referring to a different population (e.g. obese children). Two independent investigators (first and second author) screened the abstracts (k = 129) or the complete article (k = 70) of the remaining 199 publications for relevance to the defined topic and fulfillment of inclusion criteria. 2.2. Inclusion criteria (1) Studies were required to apply a randomized controlled design.

A total of 18 studies met these criteria and were included in the meta-analysis. Details of the selection process are shown in Fig. 1. 2.3. Data abstraction and validity assessment Each study was coded including information on extrinsic properties of studies, methodology, participants, treatment and methodological quality by two independent raters using a structured coding scheme (see Appendix A).1 Interrater-reliability was estimated using Cohen's Kappa for categorical variables and the 1

The full versions of the coding scheme can be requested from the authors.

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

235

Fig. 1. Flow of trials through the selection process.

intra class correlation coefficients for continuous variables. Mean Kappas for the methodology, participants and treatment characteristics were 0.98 and 0.92 for the coding of the validity. Interrater-reliabilities for single items varied between .65 and 1.00 with 96% of the items reaching a coding reliability above .75. Coding discrepancies were discussed and resolved.

the treatment effect for studies that reported means and standard deviations for intervention and control group. Hedges' g refers to the mean difference between the intervention (IG) and control group (CG) divided by the pooled standard deviation: P

P

X IG  X CG g ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : 2 2 Þ=ðn þ n ððnIG  1ÞSIG þ ðnCG  1ÞSCG IG CG  2Þ

2.4. Effect size calculation Effect sizes were calculated separately for each outcome variable, treatment group and time-point. We used Hedges' g (Hedges and Olkin, 1985) to estimate

For studies that did not report means and standard deviations on the continuous variables of interest we computed equivalent effect sizes from t and F statistics and exact probability levels (Rosenthal, 1994). For

236

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

categorical outcome data (rehospitalization, relapse, medication adherence) the logarithmized Odds Ratio (OR) was calculated and transformed into Hedges' g pffiffi (Rustenbach, 2003): g ¼ p3 lnðORÞc0:55lnðORÞ. In the case of zero in cells, we calculated the OR by raising the frequency in each cell by 0.5 (Fleiss, 1994; Haddock et al., 1998). A correction for small sample bias was applied to the effect sizes (Hedges and Olkin, 1985). 2.5. Integration of effect sizes 2.5.1. Integration of dependent effect sizes We categorized the outcome variables into four categories according to the time of assessment (post; ≤1–6-month follow-up; 7–12-month follow-up; N12month follow-up) and areas (rehospitalization and relapse; symptoms; functional outcome; knowledge; adherence). This resulted in 20 combinations, for which effect sizes were integrated. In cases, in which dependent effects remained within a category, these were combined to a total effect using the arithmetic mean (Hedges and Olkin, 1985; Matt and Cook, 1994). The inter-correlations between the outcome variables were derived from published literature or estimated. Dependent effect sizes also resulted from a study that compared two different types of PE interventions with a control group (Hornung et al., 1999a,b, 1995). In this case the mean effect size was calculated and the variances were adjusted post-hoc. 2.5.2. Integration of independent effect sizes We weighted the effect sizes by the inverse of their sampling variances (Hedges and Olkin, 1985) and calculated mean effect sizes by averaging the weighted effects within the different types of outcome categories and time-points. The homogeneity statistic Q (Shadish and Haddock, 1994) was calculated to determine whether each set of effect sizes shared a common population effect size. In the cases of homogeneity a fixed effects model (FEM) was used. In cases of heterogeneity we used a random effects model (REM).

test for a moderating effect of active control intervention versus treatment as usual, as the latter is likely to produce higher effect sizes. To identify whether patient status at baseline (inpatients versus outpatients) had an impact on relapse rates, a weighted least squares multiple regression of the individual effect sizes for relapse was conducted, controlling for treatment duration in weeks and ethnic background (European or American/Canadian versus Asian). Ethnic background was considered to be a potential moderator due to differences in health care systems and attitudes, while treatment duration was considered relevant because outpatients were receiving slightly longer treatments. The individual effect-sizes for relapse refer to those obtained in the longest follow-up period. Finally, a moderating effect of methodological quality of the studies was tested by correlating the weighted mean effect size per study (based on all effects achieved at different assessment times) with the score of the validity rating. 2.7. Sensitivity analyses A potential publication bias was examined by calculating the fail-safe N criterion (Rosenthal, 1979), which estimates how many unpublished studies with insignificant effects (dfs, defined here as 0) are needed in order to reduce the found effects to a critical effect (dcrit, defined here as 0.1), using the formula: P

kfs ¼

kð d  dcrit Þ P : dcrit  d fs

To analyze the impact of the method, we contrasted the results obtained by a fixed and a random effects model. 3. Results 3.1. Characteristics of included studies A descriptive overview of all integrated studies is presented in Table 1.

2.6. Moderator analyses In order to compare intervention formats with and without participation of family members, we used a categorical integration model. The presence of a moderating effect is indicated by heterogeneity between classes (significant Qb) and homogeneity within classes (non-significant Qw). The same procedure was used to

3.1.1. Extrinsic characteristics All studies were published between 1982 and 2005. Five studies were published before 1995 and five studies were published after 2000. Studies were conducted in Great Britain (k = 5); China (k = 4), Germany and Switzerland (k = 3), Greece (k = 1) Scandinavia (k = 2), USA and Canada (k = 3).

Table 1 Treatment Characteristics, Number of Participants (N), Effect sizes and Confidence Intervals for Relapse, Symptoms, Functioning, Knowledge and Adherence at Post (PO)- and Follow-Up (FU) Assessments Study/place

Treatment

Sample characteristics

N

FU

Effect sizes and confidence intervals Relapse

PE-P and PS vs. WL; 20 wks

Bäuml et al. (1996) Germany

PE-P-F vs. TAU; 20 wks, 8 sessions PE-P and SST vs. SR 16 wks PE-F vs. TAU 24 wks

Predominantly white UK-outpatients, schizophrenia (excluding long-stay patients) Inpatients, schizophrenia

Functioning

Knowledge

Adherence

146 P0 – FU1

–

.00 (−.32–.32) .43 (.10–.76)

–

–

163 FU1 .45 (.14–.76) FU2

–

–

–

.18 (− .13–.49) .58 (.27–.89)

–

–

Outpatients, schizophrenia, 29 P0 – non-acute, chronic (mean duration of disorder 8.9 yrs) Outpatients, schizophrenia, 61 P0 – Chien et al. (2004, 2005) China non-chronic (mean duration of 61 FU2 – disorder 2 yrs), medium education level Fries et al. (2003) PE-P and Inpatients, schizophrenia or 40 P0 – Switzerland CS vs. ST schizoaffective disorder, chronic 21 FU2 .65 a (.00–1.30) (≥2 previous admissions), non-acute Herz et al. (2000) PE-P-F vs. Outpatients, schizophrenia or 82 P0 .48 (.14–.81) USA TAU 18 months schizoaffective disorder, chronic (≥2 previous admissions), high level of education (50% college) 61 FU2 − .22 (− .73–.29) Hornung et al. (1999a,b, PE-P + LTG vs. Chronic outpatients 1995) Germany LTG; 32 wks, (≥2 previous admissions), non-acute 59 FU3 .13 (− .39–.65) 10 sessions. (≥4 wks since previous acute episode) Hornung et al. PE-P-F + LTG Chronic outpatients 61 FU2 − .12 (− .63–.39) (1999a,b, 1995) vs. LTG 32 wks, (≥2 previous admissions), non-acute 57 FU3 .24 (− .29–.77) Germany 10 sessions (≥4 wks since previous acute episode) Leavy et al. (2004) PE-F vs. TAU In-and outpatients, 42.5% first 106 FU1 .00 (− .38–.38) England 7 sessions episode psychotic disorder; 106 FU2 .69 (.30–1.08) white UK-residents Leff et al. (1982) PE-F vs. TAU High risk for relapse outpatients, 24 FU2 1.27 (.39–2.15) England schizophrenia, high EE relatives, medium education Browne et al. (1996) Ireland

− .04 (−.62–.55)

.11 (−.41–.63) .31 (− .37–.99)

– –

– –

−.36 (−.99–.28) − .88 a (− 1.54–.23) −.03 (− .65–.59) – − .14 (−1.02–.74) − .32 (− 1.20–.56) −.43 (−.1.32–.46) – –

–

.46 (.02–.90)

− .76 (− 1.20–.32)

– –

− .16 b (− .60–.27) .53 b (.09–.97)

– –

– –

– –

1.16 c (.71–1.62) .41 c (− .01–.84)

– –

– –

–

–

–

–

237

(continued on next page)

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

Atkinson et al. (1996) Scotland

Symptoms

Study/place

Treatment

Sample characteristics

N

FU

Effect sizes and confidence intervals Relapse

Li and Arthur (2005) China Merinder et al. (1999) Denmark Posner et al. (1992) Canada

Rund et al. (1994) Norway

Shin and Lukens (2002) USA

Tarrier et al. (1988) England

Tomaras et al. (2000) Greece

Xiong et al. (1994) China

Inpatients, schizophrenia, non-acute, 89 P0 ca. two thirds experiencing first admission, medium education PE-P-F vs. Inpatients, psychotic disorders, chronic 46 P0 TAU 8 wks (mean duration of disorder: 8.2 yrs) 46 FU2 55 P0 PE-F vs. WL Young (≤40) outpatients, schizophrenia; chronic 55 FU1 (≥2 previous admissions), medium to high education (mean: 12 yrs) PE-F vs. NI Recent-onset and chronic patients, 223 P0 36 wks schizophrenia, living in the community in rural China, low education (17% illiterate) 24 P0 PE-F vs. TAU Early onset of psychotic disorder 104 wks (13–18 yrs). Treatment continues from acute, inpatient to remitted, outpatient phase PE-P vs. ST Korean–American outpatients, 48 P0 10 wks chronic schizophrenia, schizoaffective or schizophreniform disorder, high education (35% college) PE-P-F vs. TAU First episode (30%) and chronic 44 FU2 2 sessions acute schizophrenia inpatients (mean duration of disorder 6.3 yrs), low to medium education PE-P-F and Non-acute outpatients with 34 P0 IPST vs. IPST; schizophrenia or schizoaffective 26 FU2 52 wks, disorder, chronic 26 FU3 13 sessions (mean duration of disorder 8.0 yrs) PE-P-F vs. TAU Urban Chinese inpatients, 63 FU1 acute schizophrenia at enrollment, 61 FU2 main treatment in outpatient status, 60 FU3 medium education (mean 9 yrs)

Symptoms

Functioning

.49 (.15–.83)

.64 (.20–1.08)

–

1.04 (.60–1.48)

.26 (− .15–.68)

– – .36 (− .16–.88)

.22 (−.37–.81) .19 (−.40–.78) – –

– – – –

– – .68 (.13–1.23) .48 e (−.15–1.12)

– – – –

1.15 (.87–1.43)

–

–

–

–

.40 (− .41–1.21) –

.89 (0.05–1.73)

–

–

–

1.11 (.50–1.91)

–

–

–

.12 (− .47–.71)

–

–

–

–

− .06 (− .61–.50) – .44 (.22–.09) – .17 (− .48–.82)

– –

– –

– – –

– – –

– – –

.60 (.16–1.03) .69 (.25–1.12) .28 −(.16–.72)

Knowledge d

Adherence

– – –

Note. PO = post-assessment, FU-1 = follow-up at ≤1–6-months; FU-2 = follow-up at 7–12-months; FU-3 = follow-up N 12-months; PE-P = psychoeducation for patients; SST = Social Skills Training; SR = supportive rehabilitation; PE-P-F = psychoeducation for patients and family; TAU = treatment as usual; PS = problem solving; WL = waiting list; CS = coping strategies; ST = supportive therapy; LTG = Leisure Time Group; PE-F = psychoeducation for family; NI = no intervention; IPST = individual psychosocial treatment. a n = 40 for this analysis. b n = 89. c n = 92. d n = 101. e n = 39.

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

Ran et al. (2003) China

PE-P-F vs. TAU 26 wks

238

Table 1 (continued)

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

3.1.2. Treatment characteristics Five studies investigated patient directed PE, six studies investigated family-directed PE and seven studies investigated patient and family-directed PE. In two studies the intervention was applied in an individual format (Leavy et al., 2004; Li and Arthur, 2005), in twelve studies PE was conducted in groups and in four studies the format was not defined. Control interventions were TAU in ten studies, WL in two studies and NSI (e.g. supportive rehabilitation, leisure groups) in six studies. Treatment was conducted in an inpatient setting in two studies, an outpatient setting in eleven studies and in a combination of both in five studies. Nine studies provided information on the duration of treatment in the patient intervention groups, with a mean2 of 27.8 weeks (SD = 18.5). Seven studies provided information on the duration of family intervention, averaging 36.8 weeks (SD = 18.4) with a mean of 11.8 sessions (SD = 14.4). The majority of studies provided pre–post-treatment data (k = = 14), eight of these additionally provided follow-up-data. Four studies provided only follow-updata. Only five studies explicitly provided data for dropout during treatment, with a mean rate of 2.6% (SD = 3.6) in the intervention groups and 3.8% (SD = 6.9) in the control groups. The rates of missing data could be reconstructed from the data provided by a further nine studies. Based on these 14 studies, the mean loss from pre to post assessment was 6.2% (SD = 15.2) in the intervention groups and 6.9% (SD = 15.6) in the control groups. 3.1.3. Patient characteristics A total of 1534 participants were investigated (mean per study = 83.7; SD = 60.1). All studies investigated patients with psychotic disorders according to the Diagnostic and Statistical Manual for Mental Disorders, the International Classification of Diseases or the Chinese Classification of Mental Disorders. Only three studies reported the use of a structured or standardized interview procedure. Two studies additionally included patients with delusional disorder, seven studies included schizoaffective disorder and four studies included other psychotic disorders. In all other cases, there were no details on diagnostic in-or exclusion criteria other than schizophrenia or psychosis. The mean age of the participants was 30.5 (SD = 4.9, k = 7). The majority

239

were men (59.3%, SD = 11.2; k = 13). The mean duration of disorder was 75.8 months (SD = 43.2; k = 5). 3.1.4. Outcome measures We categorized the outcome data into five groups, which provided the basis for the integration of effects: (1) relapse and rehospitalization data was provided by 14 studies; (2) symptoms of schizophrenia were assessed in six studies using the Brief Psychiatric Rating Scale (Chien and Chan, 2004; Fries et al., 2003; Li and Arthur, 2005; Merinder et al., 1999; Shin and Lukens, 2002), the Scale for Assessment of Positive Syndromes and the Scale for Assessment of Negative Syndromes (Browne et al., 1996; Fries et al., 2003); (3) functioning was assessed in four studies by the Global Assessment Scale (Fries et al., 2003; Hornung et al., 1995; Rund et al., 1994) and the Social Functioning Scale (Atkinson et al., 1996); (4) patients' knowledge of disorder was assessed in four studies using the Early Signs Questionnaire (Herz et al., 2000), a self-devised measure to assess knowledge of schizophrenia (Merinder et al., 1999) and the Knowledge about Schizophrenia Interview (Li and Arthur, 2005). In the study by Fries et al. (2003) the instrument was not specified; (5) medication adherence was assessed in four studies by dichotomizing the continuity of medication intake. 3.1.5. Validity The maximum possible score from the rating of validity according to the coding plan (see Appendix A) was 18, indicating maximum validity and zero indicating minimum validity. Values for the included studies ranged from 15 (Herz et al., 2000; Ran et al., 2003) to 6 (Atkinson et al., 1996). The mean score across studies was 11.0 (SD = 2.5), which speaks for a medium validity of the included studies. Specifically, 16 studies made use of multiple measurement methods, 15 studies used reliable instruments, 12 studies provided information that allowed to conclude that the intervention groups were comparable with regard to socio-demographic properties or clinical variables, six studies provided information on inter-rater reliability of the measures, six studies used blinded raters, five studies used trained raters to do the assessments, five studies varied the treatment setting and four studies used multiple recruiting methods. 3.2. Efficacy of PE for different types of outcome variables and time-points

2 Means of patient and study characteristics were calculated by sample-size-weighted averaging of means or numbers reported in the primary studies.

The effect size calculation for the five outcome areas and four follow-up time points resulted in a total of 49

240

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

Table 2 Mean weighted effect sizes from controlled comparisons as a function of outcome category and follow-up period

3.3. Impact of treatment format, patient status and validity

Outcome category

In an additional analysis we calculated mean effect sizes separately for interventions directed at family members and patients versus interventions directed solely at patients. Comparisons were restricted to outcome measures and time points that were reported in at least two studies within each format. This was the case for symptoms at post-treatment and relapse/ rehospitalization at 7–12 month follow-up. Table 3 depicts the results of these analyses. With regard to symptom change, the average effect sizes for both treatment modalities were small and insignificant. However, there were significant differences between the modalities as indicated by the between group homogeneity statistic (Qb = 10.6, p = .00, power = .90), while the effect sizes were homogeneous within the groups (non-significant Qw). Small effects were also achieved for relapse/rehospitalization at 7–12 month follow-up, but only the effect for the family included intervention was significant. Here too, the differences between the modalities were significant (Qb = 11.4, p = .00, power = .92) while the effect sizes were homogeneous within the groups. In sum, although family PE was superior, neither format had a significant effect on symptom change. However, only family focused interventions were able to significantly reduce the rates of relapse and rehospitalization. We tested for a moderating effect of control-group by comparing the effect sizes for rehospitalization and relapse achieved for interventions that were compared to TAU (not including wait-list-conditions) with those achieved for interventions that were compared to NSI.

Post-assessment Relapse/ Rehospitalization a Symptoms a Functional outcome a Knowledge a Medication adherence a Follow-up ≤6 months Relapse/ Rehospitalization Follow-up 7–12 months Relapse/ Rehospitalization a Symptoms Functional outcome a Follow-upN12 months Relapse/ Rehospitalization

k N 5 452 6 3 4 2

d

95% CI 0.53

p

Q

0.12–0.95 0.01 3.45

313 0.29 −0.13–0.70 210 − 0.03 −0.84–0.78 278 0.48 0.12–0.83 171 − 0.25 −1.25–0.75

0.08 0.97 0.00 0.31

5.38 2.82 3.31 1.00

4 387

0.35

0.14–0.55 0.00 4.60

7 362

0.48

0.15–0.82 0.00 5.93

3 128 2 112

0.19 −0.16–0.55 0.14 0.75 0.19 −0.59–0.97 0.32 1.00

3 144

0.21 −0.07–0.49 0.07 0.15

Note. k = number of integrated studies; N = number of investigated patients, d = weighted mean effect size; CI = confidence interval for d; p = level of significance for d; Q = homogeneity statistic for d. a Integrations carried out in random effects model.

effect sizes as presented in Table 1. These effect sizes ranged from − .88 (functioning, Fries et al., 2003) to 1.27 (relapse, Leff et al., 1982). Due to the small number of studies and the fact that most studies did not provide data for all outcome areas and time-points an integration of effect sizes was not possible for each combination of outcome area and time point. As shown in Table 2, the weighted effect sizes for relapse and rehospitalization ranged from medium (posttreatment) to small (follow-up ≤6 months and 7– 12 months) and were no longer significant at follow-up N12 months. Confidence intervals were generally large. General efficacy of PE for symptom reduction could be calculated for post-treatment and follow-up 7–12 months. Both effect sizes were small and failed significance. Effect sizes for functional outcome could be calculated for post-treatment and the follow-up period 7–12 months. Both effect sizes were not significant. An integrated effect size for a gain in knowledge about the disorder could only be calculated for postassessment. This effect size was small, but in the range of small to large effect sizes. Finally, an integrated effect size for the effect of PE on medication adherence could only be calculated for post-assessment. This effect size was small, negative and insignificant.

Table 3 Analysis of variance in effect sizes as a function of treatment format Treatment modality

k N

Random effects model d

Symptoms at post-assessment PE without family PE with family Relapse/ Rehospitalizations at 7–12 month-follow-up PE without family PE with family

95% CI

p

Qwi (p)

3 117 0.24 − .39–.86 .23 3.86 (.15) 3 196 0.33 − .26–.93 .14 0.58 (.75)

2 101 0.18 − .47–.82 .30 1.72 (.19) 6 322 0.48 .10–.85 .00 4.78 (.44)

Note. k = number of integrated studies; N = number of investigated patients, d weighted mean effect size; CI = confidence interval for d; p = level of significance for d; Qwi = homogeneity statistic within each group.

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

241

Sufficient numbers of studies were available for comparisons at post and follow-up-2. At post, studies with a NSI control condition (k = 2) produced a significantly higher effect for relapse and rehospitalization than studies with a TAU control condition (k = 3) (d¯NSI = 0.64, d¯TAU = 0.46, Qb = 17.1, p = .00, power = .99) while the effect sizes were homogeneous within the groups (QwNSI = 3.8, n.s.; QwTAU = 0.0, n.s.). At followup studies with a NSI control condition (k = 3) produced a significantly smaller effect for relapse and rehospitalization than studies with a TAU control condition (k = 4) ( d¯ N S I = 0.24, d¯ TA U = 0.33, Q b = 13.6, p = .00, power = .96) while the effect sizes were homogeneous within the groups (QwNSI = 2.0, n.s.; QwTAU = 1.6, n.s.). Taken together, there was no clear trend for higher effects of NSI compared to TAU control conditions. Patient status was not a significant predictor of relapse after controlling for duration and ethnic background in the regression model (F = 2.58, p = .12). However there was a tendency for outpatients to profit more (β = − .44, p = .14). The methodological quality of the studies as assessed by the validity rating was not significantly correlated with the weighted mean effect size per study (r = −.27, p = .28, k = 18).

various outcome domains, using clearly defined inclusion criteria with regard to the PE-intervention. Due to this criterion, a large number of studies encountered in the literature search and included in other analyses were excluded. For example, of the 25 studies in the analysis by Pitschel-Walz (Pitschel-Walz et al., 2001) only three studies fulfilled our criteria, the most frequent reasons for exclusion being that the interventions did not fulfill our criteria for PE or diagnosis. Overall, PE was shown to have a short term medium effect on relapse, which grew slightly smaller with time and was no longer significant after one year. PE had a small, but significant effect on knowledge by the end of the intervention. The effects of PE on symptom reduction, functioning and medication adherence were insignificant. The format comparisons favored PE directed at patients and their family over a mere patient intervention. Furthermore, PE proved significant with regard to relapse when the family was included while all effect sizes for PE directed at patients only were insignificant. The differences between PE compared with treatment-as-usual and PE compared to nonspecific control interventions were inconsistent.

3.4. Results of the sensitivity analyses

Due to the heterogeneity of studies and the low numbers for each outcome category, confidence intervals were generally large. As a consequence, the effect sizes can not reliably be categorized as small or medium according to the classification by Cohen (1977). The fact that methodologically weaker studies did not produce significantly differing effects underlines the validity of our findings. However, the numbers of fail safe studies needed to reduce some of the effect sizes to insignificance were low. It is possible that there are a row of unpublished studies with zero effects and that the overall effect sizes might be overestimated in this analysis. Although the ES size-differences between nonspecific-control-group comparisons and TAU or WL conditions were not consistently in favor of inactive control-interventions, it must be noted that the majority of patient directed interventions (80%) and only a minority of family directed interventions (14%) employed an active control-condition and that this might have contributed to effect size-differences between the intervention formats. The fact that the included studies were conducted in a variety of different countries and cultures can be considered as a strength of the current meta-analysis with regard to the generalizability of the findings.

3.4.1. Publication bias We calculated the fail safe N separately for each category of integrated effect sizes that produced significant effect sizes. The fail-safe numbers were as follows: 22 (relapse at posttreatment), 15 (knowledge at post-treatment), 10 (relapse at follow-up-1) and 27 (relapse at follow-up-2). 3.4.2. Comparison of integration methods Three out of ten integrations produced homogeneous effects and were carried out using the fixed effects model. In order to test whether the results are independent of the selected integration method, we compared the results of these analyses to results we would have obtained using a random effects model. In all the integrations the effect sizes were independent of the model used. 4. Discussion 4.1. Summary of results The aim of this analysis was to establish the efficacy of PE with and without family inclusion with regard to

4.2. Generalizability

242

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

4.3. Discussion of findings The finding that there is a small effect on relapse and rehospitalization for PE directed at patients and family members is in line with the other meta-analyses on PE (Pekkala and Merinder, 2004; Pitschel-Walz et al., 2001). The size of the effect is comparable with overall effect sizes found for other forms of effective interventions in schizophrenia, such as CBT (Jones et al., 2004; Pilling et al., 2002a,b), antipsychotic medication (Leucht et al., 1999) or case-management (Ziguras and Stuart, 2000) which are generally small, presumably due to multifaceted etiology and maintenance factors and the large heterogeneity of symptoms in schizophrenia. The finding that PE was ineffective in improving adherence is in line with the systematic review by Zygmunt et al. (2002) who also found interventions and family programs relying on PE to improve medication adherence to be ineffective. Divergence to results from older reviews and analyses might be explained by the diversity of methods to assess adherence. Future studies could profit from following the suggestions for assessing adherence put forward by Velligan et al. (2006). The most interesting finding is that PE offered solely to patients was ineffective. Although PE directed at patients during inpatient stay is frequently employed and has practical advantages (Rummel-Kluge et al., 2006), there are also some relevant concerns with it which might account for the insignificance. First, many patients with schizophrenia have little insight into the presence of a mental disorder or do not attribute symptoms to it (Gharabawi et al., 2006; Sevy et al., 2004). Many of these patients do not readily accept the alternative illness interpretation put forward in PE. Reluctance is often particularly pronounced if patients feel that their individual and often psychosocial explanations are not being taken seriously (Hofer et al., 2001). Second, due to the large diversity of symptoms in schizophrenia, many patients find it hard to detect their own experiences in the standardized program descriptions of symptoms and may conclude that the diagnosis does not apply to them. Third, although PE programs have become more interactive there is still a lot of directive lecturing involved, which is problematic for patients suffering from poor concentration and memory and might explain why more interactive presentation of information as well as more behavioral components have been found to produce higher effects (Buttner, 1996; Dixon and Lehman, 1995; Mullen et al., 1985; Zygmunt et al., 2002). Finally, the programs do not require to accompany patients after the group has terminated and support them in integrating the knowledge into their daily life (e.g. monitoring medication,

stress and early signs of relapse). For an intervention to have an impact, the basic messages have to be understood and accepted by the patient, related to his or her own problems as well as remembered and integrated into everyday life. PE offered to patients alone during inpatient stay seems vulnerable with regard to any of these steps. Some of the listed problems might be reduced by delivering PE to patients in an individual format, however none of the patient interventions investigated in this analysis has taken this approach. Also, as the moderator analysis revealed a tendency for outpatients to improve more from PE with regard to relapse, this could also prove to be a more promising approach that warrants further research. We assume that the inclusion of the family might automatically resolve some of the difficulties listed above as family members are involved in integrating the knowledge into reality and supporting patients in the monitoring of symptoms and stress level and that this might explain why it is more effective. In addition, family directed PE might reduce negative communication by promoting understanding (Szmukler et al., 1996) and thus contribute to relapse prevention, even without explicit communication training. The relevance of family inclusion for effective PE might also explain why it was effective for relapse, while symptoms remained stable. Possibly the main effect of PE is that family members learn to cope better with schizophrenia, resulting in less rehospitalization, in spite of unchanged symptoms. While PE patient groups are frequently employed, the inclusion of relatives is still not an integral part of treatment for schizophrenia, despite its proven effectiveness in reducing medical costs (Mino et al., 2007). A more detailed analysis resulting from the investigation by Rummel-Kluge et al. (2006) showed that only 2% of the family members had taken part in PE in the investigated year. Dixon (1999) found that only 8% of all questioned patients reported that their relatives had ever taken part in supportive or PE interventions. Comparable percentages have been found for Italy (Magliano et al., 2006). Barriers to the implementation of family PE are mainly seen to come from organizational factors (Dixon, 1999). In particular, the caseloads of many clinicians in psychiatric institutions often do not leave time for the additional effort of contacting and motivating relatives to participate. 4.4. Conclusions Although common in clinical practice, PE groups directed solely at patients are under-researched and remain to be proven effective. The additional effort of integrating relatives into treatment is definitely worth-

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

while and should be encouraged where possible. As family inclusion is not always feasible it could be helpful to further investigate the barriers to effectiveness of patient directed PE, which might improve if it is offered to patients who are out of the acute stage and if patients are supported individually in transferring their knowledge into everyday life. Role of Funding Source There was no specific funding for this study. Costs for literature were covered by the Department of Clinical Psychology at the Philipps-University Marburg. Contributors The authors designed the study, analyzed the data and wrote the manuscript. Conflict of Interest All authors declare that they have no conflict of interest. Acknowledgement We would like to thank Christiane Suttner for coding part of the studies and Prof. H-H. Schulze for methodological advice.

Appendix A Coding scheme for study characteristics (1) Identification (authors, year of publication, place of study); (2) methodology (recruitment methods, type of outcome variables [symptoms, relapse, functional outcome, adherence and knowledge about disorder], measurement points, diagnostic procedure and classification system, protocol and control of medication); (3) participants (number, diagnosis, status [chronic, acute], duration of disorder, age, gender); (4) treatment (setting, number of sessions for each type of intervention, kind of additional interventions, intervention in control group, focus [patients, relatives or combination], method [interactive versus directive], format [group, single, combined], duration of intervention in weeks, therapists [psychologists, doctors, nurses] and dropout-rate). Coding scheme for validity rating (1) Validity of the statistical conclusion (comparability of intervention groups with regard to socio-demographic or clinical variables, reliability of measures); (2) internal validity (blindness of investigators, assessment of interrater-reliability, training of raters); (3) construct validity (multiple outcome measures, one intervention: PE) and (4) external validity (multiple recruitment methods, multiple settings, heterogeneous patients within each group). Studies were coded 2 if the criterion was fulfilled, 1 if partially fulfilled and 0 if not fulfilled or not reported. The maximum possible score was 18.

243

References Atkinson, J.M., Coia, D.A., Gilmour, W.H., Harper, J.P., 1996. The impact of education groups for people with schizophrenia on social functioning and quality of life. Br. J. Psychiatry 168, 199–204. Barbato, A., D`Avanzo, B., 2000. Family intervenitons in schizophrenia and related disorders: a critical review. Acta Psychiatr. Scand. 102, 81–97. Bäuml, J., Kissling, W., Pitschel-Walz, G., 1996. Psychoedukative Gruppen für schizophrene Patienten: Einfluss auf Wissensstand und Compliance. Ergebnisse der Münchener PIP-Studie. Nervenheilkunde 15, 145–150. Browne, S., Roe, M., Lane, A., Gervin, M., Morris, M., Kinsella, A., et al., 1996. A preliminary report on the effect of a psychosocial and educative rehabilitation program on quality of life and symptomatology in schizophrenia. Eur. Psychiatr. 11, 386–389. Buttner, P., 1996. Die Wirksamkeit psychoedukativer Verfahren in der Schizophreniebehandlung. In: Stark, A. (Ed.), Verhaltenstherapeutische und psychoedukative Ansätze im Umgang mit schizophren Erkrankten. Tübingen, Dgvt-Verlag, pp. 193–206. Chien, W.T., Chan, S., 2004. One-year follow-up of a multiple-familygroup intervention for Chinese families of patients with schizophrenia. Psychiatr. Serv. 55, 1276–1284. Chien, W.T.S.C., Morrisey, J., Thompson, D., 2005. Effectiveness of a mutual support group for families of patients with schizophrenia. J. Adv. Nurs. 51, 595–608. Cohen, J., 1977. Statistical Power Analysis for the Behavioural Science. Academic Press, New York. Dixon, L., 1999. Providing services to families of persons with schizophrenia: present and future. J. Mental Health Policy Econ. 2, 3–8. Dixon, L., Lehman, A.F., 1995. Family interventions for schizophrenia. Schizophr. Bull. 21, 631–643. Fleiss, J.L., 1994. Measures of effect size for categorical data. In: Cooper, H., Hedges, L.V. (Eds.), The Handbook of Research Synthesis. Russel Sage, New York, pp. 245–260. Fries, A., Pfammatter, M., Andres, K., Brenner, H.D., 2003. Wirksamkeit und Prozessmerkmale einer psychoedukativen und bewältigungsorientierten Gruppentherapie für schizophren und schizoaffektiv Erkrankte. Verhaltenstherapie 13, 237–243. Gharabawi, G.M., Lasser, R.A., Bossie, C.A., Zhu, Y., Amador, X., 2006. Insight and its relationship to clinical outcomes in patients with schizophrenia or schizoaffective disorder receiving longacting risperidone. Int. Clin. Psychopharmacol. 21, 233–240. Goulet, J., Lalonde, P., Lavoie, G., Jodoin, F., 1993. Effets d'une éducation au traitement neuroleptique chez de jeunes psychotiques. Can. J. Psychiatry 38, 571–573. Haddock, C.K., Rindskopf, D., Shadish, W.R., 1998. Using odds ratios as effect sizes for meta-analysis of dichotomous data. A primer on methods and issues. Psychol. Methods 3, 339–353. Hedges, L.V., Olkin, I., 1985. Statistical Methods for Meta-analysis. Academic Press, London. Herz, M.I., Lamberti, J.S., Mintz, J., Scott, R., O'Dell, S.P., McCartan, L., et al., 2000. A program for relapse prevention in schizophrenia: a controlled study. Arch. Gen. Psychiatry 57, 277–283. Hofer, E., Amering, M., Windhaber, J., Wancater, M., Eberstaller, M., 2001. Does psychoeducation change the concept of illness of schizophrenic patients. Psychiatr. Prax. 28, 287–291. Hogarty, G., Sander, J.K., Greenwald, D., DiBarry, A.L., Cooley, S., Ulrich, M.S., et al., 1997. Three-year trials of personal therapy among schizophrenic patients living with or independent of family, I: description of study and effects on relapse rates. Am. J. Psychiatry 154, 1504–1513.

244

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245

Hornung, W.P., Holle, R., Schulze Mönking, H., Klingberg, S., Buchkremer, G., 1995. Psychoedukativ-psychotherapeutische Behandlung von schizophrenen Patienten und ihren Bezugspersonen. Ergebnisse einer 1-Jahres-Katamnese. Nervenarzt 66, 828–834. Hornung, W.P., Feldmann, R., Klingberg, S., Buchkremer, G., Reker, T., 1999a. Long-term effects of a psychoeducational psychotherapeutic intervention for schizophrenic outpatients and their keypersons — results of a five-year follow-up. Eur. Arch. Psychiatry Clin. Neurosci. 249, 162–167. Hornung, W.P., Feldmann, R., Schonauer, K., Schäfer, A., Mönking, H.S., Klingberg, S., et al., 1999b. Psychoedukativ-psychotherapeutische Behandlung von schizophrenen Patienten und ihren Bezugspersonen; II. Ergänzende Befunde der 2-Jahres-Katamnese. Nervenarzt 70, 444–449. Jones, C., Cormac, I., Silveira da Mota Neto, J.I., Campbell, C., 2004. Cognitive therapy for schizophrenia. Cochrane Database Syst. Rev. 4. Leavy, G., Gulamhussein, S., Papadopoulos, C., Johnson-Sabine, E., Blizard, B., King, M., 2004. A randomized controlled trial of a brief intervention for families with a first episode of psychosis. Psychol. Med. 34, 423–431. Leff, J., Kuipers, L., Berkowitz, R., Eberlein-Fries, R., Sturgeon, D., 1982. A controlled trial of social interventions in the families of schizophrenic patients. Br. J. Psychiatry 141, 121–134. Leucht, S., Pitschel-Walz, G., Abraham, D., Kissling, W., 1999. Efficacy and extrapyramidal side-effects of the new antipsychotics olanzapine, quetiapine, risperidone, and sertindole compared to conventional antipsychotics and placebo. A meta-analysis of randomized controlled trials. Schizophr. Res. 35, 51–68. Li, Z., Arthur, D., 2005. Family education for people with schizophrenia in Beijing, China. Br. J. Psychiatry 187, 339–345. Magliano, L., Fiorillo, A., Malangone, C., De Rosa, C., Maj, M., Family Intervention, W., 2006. Implementing psychoeducational interventions in Italy for patients with schizophrenia and their families. Psychiatr. Serv. 57, 266–269. Mari, J.J., Streiner, D.L., 1994. An overview of family interventions and relapse on schizophrenia: meta-analysis of research findings. Psychol. Med. 24, 565–578. Matt, G.E., Cook, T.D., 1994. Threats to validity of research synthesis. In: Cooper, H., Hedges, L.V. (Eds.), The Handbook of Research Synthesis. Russel Sage, New York, pp. 503–520. Merinder, L.B., 2000. Patient education in schizophrenia: a review. Acta Psychiatr. Scand. 102, 98–106. Merinder, L.B., Viuff, A.G., Laugesen, H.D., Clemmensen, K., Misfelt, S., Espensen, B., 1999. Patient and relative education in community psychiatry: a randomised controlled trial regarding its effectiveness. Soc. Psychiatry Psychiatr. Epidemiol. 34, 287–294. Mino, Y., Shimodera, S., Inoue, S., Fujita, H., Fukuzawa, K., 2007. Medical cost analysis of family psychoeducation for schizophrenia. Psychiatry Clin. Neurosci. 61, 20–24. Mueser, K.T., Corrigan, P.W., Hilton, D.W., Tanzman, B., Schaub, A., Gingerich, S., et al., 2002. Illness management and recovery: a review of the research. Psychiatr. Serv. 53, 1272–1284. Mullen, P.D., Green, L.W., Persinger, G.S., 1985. Clinical trials of patient education for chronic conditions: a comparative metaanalysis of intervention types. Prev. Med. 14, 753–781. Pekkala, E., Merinder, L., 2004. Psychoeducation for schizophrenia (Cochrane Review). The Cochrane Library, Issue 1. John Wiley and Sons, Ltd, Chichester. Pfammatter, M., Junghan, U.M., Brenner, H.D., 2006. Efficacy of psychological therapy in schizophrenia: conclusions from metaanalyses. Schizophr. Bull. 32, 64–80.

Pharoa, F., Mari, J., Rathbone, J., Wong, W., 2006. Family intervention for schizophrenia. Cochrane Database Sys Rev, Issue 4, CD000088. Pilling, S., Bebbington, P., Kuipers, E., Garety, P., Geddes, J., Orbach, G., et al., 2002a. Psychological treatments in schizophrenia: I. Metaanalysis of family interventions and cognitive behaviour therapy. Psychol. Med. 31, 763–782. Pilling, S., Bebbington, P., Kuipers, E., Garety, P., Geddes, J., Martindale, B., Orbach, G., Morgan, C., 2002b. Psychological treatments in schizophrenia: II. Meta-analysis of randomised controlled trials of social skills training and cognitive remediation. Psychol. Medline 32, 783–791. Pitschel-Walz, G., Leucht, S., Bäuml, J., Kissling, W., Engel, R.R., 2001. The effect of family interventions on relapse and rehospitalization in schizophrenia — a meta-analysis. Schizophr. Bull. 27, 73–92. Posner, C.M., Wilson, K.G., Kral, M.J., Lander, S., Mellwraith, R.D., 1992. Family psychoeducational support groups in schizophrenia. Am. J. Orthopsychiatr. 62, 206–218. Ran, M.S., Xiang, M.Z., Chan, C.L., Leff, J., Simpson, P., Huang, M.S., et al., 2003. Effectiveness of psychoeducational intervention for rural Chinese families experiencing schizophrenia — a randomised controlled trial. Soc. Psychiatry Psychiatr. Epidemiol. 38, 69–75. Randolph, E., Spencer, E.T.C., Glynn, S.M., Paz, G.G., Leong, G.B., Shander, A.L., et al., 1994. Behavioural family management in schizophrenia. Outcome of a clinic-based intervention. Br. J. Psychiatry 164, 501–506. Razali, M.S., Yahya, H., 1995. Comliance with treatment in schizophrenia: a drug intervention program in a developing country. Acta Psychiatr. Scand. 91, 331–335. Rosenthal, R., 1979. The “file-drawer-problem” and tolerance of null results. Psychol. Bull. 86, 638–641. Rosenthal, R., 1994. The fugitive literature. In: Cooper, H., Hedges, L.V. (Eds.), The Handbook of Research Synthesis. Russel Sage, New York, pp. 85–94. Rummel-Kluge, C., Pitschel-Walz, G., Bäuml, J., Kissling, W., 2006. Psychoedcation in schizophrenia — results of a survey of all psychiatric institutions in Germany, Austria, and Switzerland. Schizophr. Bull. 32, 765–775. Rund, B.-R., Moe, L., Sollien, T., Fjell, A., 1994. The Psychosis Project: outcome and cost-effectiveness of a psychoeducational treatment programme for schizophrenic adolescents. Acta Psychiatr. Scand. 89, 211–218. Rustenbach, S.J., 2003. Metaanalyse. Eine anwendungsorientierte Einführung. Hans Huber, Bern. Sevy, S., Nathanson, K., Visweswaraiah, H., Amador, X., 2004. The relationship between insight and symptoms in schizophrenia. Compr. Psychiatry 45, 16–19. Shadish, W.R., Haddock, C.K., 1994. Combining estimates of effects size. In: Cooper, H., Hedges, L.V. (Eds.), The Handbook of Research Synthesis. Russel Sage, New York, pp. 261–281. Shin, S.K., Lukens, E.P., 2002. Effects of psychoeducation for Korean Americans with chronic mental illness. Psychiatr. Serv. 53, 1125–1131. Szmukler, G.I., Herman, H., Colusa, S., Benson, A., Bloch, S., 1996. A controlled trial of a counselling intervention for caregivers of relatives with schizophrenia. Soc. Psychiatry Psychiatr. Epidemiol. 31, 149–155. Tarrier, N., Barrowclough, C., Vaugh, C., Bamrah, J.S., Porceddu, K., Watts, S., et al., 1988. Community management schizophrenia: a controlled trial of behavioral interventions with families to reduce relapse. Br. J. Psychiatry 153, 532–542. Tomaras, V., Mavreas, V., Economou, M., Ioannovich, E., Kardydi, V., Stefanis, C.T., 2000. The effect of family intervention on chronic

T.M. Lincoln et al. / Schizophrenia Research 96 (2007) 232–245 schizophrenics under individual psychosocial treatment: a 3-year study. Soc. Psychiatry Psychiatr. Epidemiol. 35, 487–493. Vaughan, K., Doyle, M., McConaghy, N., Blaszczynski, A., Fox, A., Tarrier, N., 1992. The Sydney intervention trial: a controlled trial of relatives' counseling to reduce schizophrenic relapse. Soc. Psychiatry Psychiatr. Epidemiol. 27, 16–21. Velligan, D.I., Lam, Y.-W.F., Glahn, D.C., Barrett, J.A., Maples, N.J., Ereshefsky, L., et al., 2006. Defining and assessing adherence to oral antipsychotics: a review of the literature. Schizophr. Bull. 32, 724–742. Wiedemann, G., Klingberg, S., Pitschel-Walz, G., Psychoedukation, A., 2003. Psychoedukative Interventionen in der Behandlung von Patienten mit schizophrenen Störungen. Nervenarzt 74, 789–808.

245

Wunderlich, U., Wiedemann, G., Buchkremer, G., 1996. Sind psychosoziale Interventionen bei schizophrenen Patienten wirksam? Eine Metaanalyse. Verhaltenstherapie 6, 4–13. Xiong, W., Phillips, M.R., Hu, X., Wang, R., Dai, Q., Kleinman, J., et al., 1994. Family based intervention for schizophrenic patients in China: a randomised controlled trial. Br. J. Psychiatry 165, 239–247. Ziguras, S.J., Stuart, G.W., 2000. A meta-analysis of the effectiveness of mental health case management over 20 years. Psychiatr. Serv. 51, 1410–1421. Zygmunt, A., Olfson, M., Boyer, C.A., Mechanic, D., 2002. Interventions to improve medication adherence in schizophrenia. Am. J. Psychiatry 159, 1653–1664.