Types of review methods

There are several kinds of research review methods. The most basic is the simple literature review that only summarizes the literature on a given topic in narrative form and does not necessarily follow rigorous and explicit strategies for search, selection, or analysis of the literature [7]. A more in-depth research review method is the systematic review that implements an explicit method for conducting a comprehensive search, and for selecting and evaluating previous quantitative studies on a well-defined clinical question (e.g., does PD reduce youth depression?). Systematic reviews include narrative analysis of the primary studies and, sometimes, descriptive statistics and quantification of qualitative information. They provide reliable evidence on gaps in current research, solid grounds for determining evidence-based practice, and the first stage of meta-analysis studies [8]. While systematic reviews have traditionally focused predominantly on questions relating to the effectiveness of interventions, there are at least 10 different types of systematic reviews that address diverse questions and the “information needs of healthcare professionals and policy makers” [9]. One type, which is particularly pertinent to this study, is a methodological systematic review that examines methodological issues relating, for example, to design and conduct. A meta-analysis is considered the highest level of evidence and the gold standard of literature-based research, where carefully selected data from primary quantitative studies are combined with rigorous statistical procedures to calculate an overall effect size [10].

Nevertheless, given the relatively small number of PD studies, we implemented an integrative review approach that allows for the combination of publications with diverse methodologies including quantitative, qualitative, and mixed methods [11]. This review approach can produce a more comprehensive portrayal of research trends in a given field. As such, integrative reviews can help generate new knowledge, as well as inform research, practice, and policy initiatives [12].

The present study

Drawing on a recent typology of systematic reviews in the medical and health sciences, the present study is an integrative methodological systematic review [9]. Hence the overarching question in this review was to determine which trends and methodological issues could be identified in psychodrama psychotherapy research. Accordingly, this systematic review had two aims: (a) to summarize 10 years of PD intervention research with all types of participants, interventions, comparisons, and outcomes according to the eligibility criteria described below; and (b) to critically evaluate methodological issues to help determine the requirements for future studies.

The present study is part of a wider project reviewing both PD and drama therapy research [13]. But because these two treatment modalities are considered different healthcare professions, two distinct reports were generated. Briefly, in psychodrama, clients typically use role-play to enact themselves, parts of themselves, or significant others in their real lives, and hence work more directly on reality-based issues. In contrast, drama therapy is more fantasy-based and clients typically use role-play to enact fictional and symbolic roles, use storytelling, puppetry, masks, and miniature objects to work more indirectly and with greater dramatic distance from their issues. However, some contemporary variants of psychodrama practice involve working with metaphors and imagination, thus blurring the boundaries between classical PD and drama therapy [14–16].

Data extraction

A spreadsheet for data extraction was piloted and then used to record information of interest and the PICOS framework specifying participants, interventions, comparisons, outcomes, and study designs [17]. Information was extracted from each study on: (1) study method: quantitative, qualitative, mixed methods; (2) study design (e.g., randomized / nonrandomized group comparison pre-post, single group pre-post, single subject design, and follow up); (3) characteristics of participants: N, age, gender, diagnosis, setting, country; (4) intervention type (duration, frequency, structure, protocol) versus control or placebo; (5) primary outcome measures; (6) results (significance, effect size). The authors of five studies were contacted with questions to verify information about their study [18–23].

Eligibility criteria

Eligibility criteria were for the publication to be an intervention research with quantitative and/or qualitative data, written in English, and published in a peer reviewed journal. No restrictions were imposed on publication status. Publication date was restricted to between January 2007 and December 2017 (the date last searched was 10 January 2018). Eligibility assessment was performed independently by the two authors and cases of disagreement were discussed until all issues were resolved and there was consensus.

Assessment of qualitative and mixed methods studies

The assessment of the qualitative and mixed methods studies drew on Hawker and colleagues’ tool for critically and systematically reviewing research conducted using different paradigms [24]. Consistent with the purpose of this review, we focused on the seven methodological dimensions defined by Hawker et al.: method and data, sampling, data analysis, ethics and bias, findings, and transferability. Three dimensions that do not address methodological issues are not reported in this review (i.e., “abstract and title,” “introduction and aims,” “implications and usefulness”).

Risk of bias assessment

Consistent with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [25], we used the Risk of Bias (RoB) assessment tool that was developed by the Cochrane Collaboration for randomized controlled trials [26]. Generally, RoB assessment focuses on the internal validity of a study, and in particular on methodological flaws that can lead to bias. Because the results of a study may in fact be unbiased despite a methodological flaw, the tool assesses the risk of bias. Bias is assessed in several domains for which there is empirical evidence that biasing can influence the estimates of an intervention’s effectiveness. A bias can lead to underestimation or overestimation of the true intervention effect [25].

Recently, it has been noted that because the RoB tool was developed within the field of medicine, its application to the context of psychotherapy outcome research requires several adjustments [27]. Drawing on the adjusted criteria suggested in Munder and Barth, we assessed the following six potential sources of bias: (1) random sequence generation, (2) concealment of allocation to conditions, (3) blinding of participants and personnel to condition assignment, (4) handling of incomplete outcome data, (5) selective outcome reporting, and (6) treatment implementation. For each source of bias, two raters judged whether the risk of bias was low, unclear, or high according to definitions and examples in Munder and Barth (2018). Note that because all the primary outcomes in the studies were self-reported and because in psychotherapy it is impossible to blind patients to intervention aims and content we did not code RoB given the absence of blinding of the outcome assessors [28]. Hence, the RoB tool was only used for the assessment of studies with a randomized group comparison design.

General characteristics of psychodrama studies

As can be seen in Fig 2, the majority of the PD research publications were from Turkey (n = 12, 39%), with the next largest groups from Italy (n = 4, 13%) and Israel (n = 3, 10%), and USA (n = 3, 10%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Frequency of included publications by country.

https://doi.org/10.1371/journal.pone.0212575.g002

The frequency of PD publications per year ranged from one (in 2008) to eight (26% in 2017), with a total average of three publications per year. Of the 31 PD publications, 20 were quantitative, four were qualitative, and seven had a mixed methods design.

Regarding client populations and primary outcomes, as can be seen in Fig 3, the majority of the PD research publications dealt with youth at risk (n = 10, 32%), followed by students (n = 4, 13%), and adults with different mental health conditions (n = 4, 13%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Primary outcomes.

Frequency of primary outcomes. SWB = subjective wellbeing, GSF = general symptoms and functioning, BP = behavioral problems.

https://doi.org/10.1371/journal.pone.0212575.g003

The most frequent outcomes, in descending order, were adolescents behavioral problems (n = 6, 19%), followed by anxiety (n = 5, 16%) and depression (n = 4, 13%). Next was quality of life (n = 3, 10%) followed by symptoms and global functioning (n = 3, 10%). Two studies measured attachment (6%) and two studies measured in-session process variables (6%), of which one measured dramatic engagement [29]. Only one study measured spontaneity, which is a core construct in PD theory [30]. As also seen in Fig 3, other primary outcomes included subjective well-being, smoking cessation, burnout, hopelessness, loneliness, sense of coherence, public stigma, self-stigma, self-esteem, self-efficacy, and self-concept. The next sections present narrative reviews of the characteristics of these studies as a function of methodology.

Qualitative studies

All four qualitative studies had a single group design. As seen in Table 1, the number of participants in the qualitative studies ranged from 7 to 13 (M = 10, SD = 2.6). All the studies involved adults: two studies examined Italian adults with cancer [20, 31], one study dealt with adults with mental health conditions in the USA [32], and one study was about Turkish mothers of children with Attention Deficit Hyperactivity Disorder (ADHD) [33]. Two of the studies dealt with hospital settings [31, 32], one study was conducted in a special education center [33], and one at a university [20].

PD treatment was delivered in a group format in all four studies: one study delivered analytic PD group therapy [20] and the three others followed the “classical” PD approach. In three studies, session duration ranged from 1.5 to 3 hours per week, and the number of sessions ranged from 1 to 80. The exception is the study with the analytic PD group that delivered two 1.5 hours sessions per month, for two years [20]. In terms of session structure, two studies had a structured [32] or a semi-structured format [33], one followed the standard three PD phases of warm-up, action, and sharing [31], and one followed an analytic 3-phase structure that included free associations as the group warm-up phase [20]. Information concerning the qualifications of the therapists was only provided in two studies: “a drama therapist educated in a national psychodrama institution and an assistant therapist” [33] and a “certified psychodrama therapist” [32]. Only one study indicated that the PD therapist received clinical supervision during treatment implementation [33].

Regarding data collection, two studies conducted post-treatment interviews [20, 31], and one study included both post-treatment interviews and direct observation during the sessions [32]. In one study, data were only collected by notes taken during the sessions [33]. Data analysis involved content/ thematic analysis [20, 32], interpretative phenomenological analysis [31], and analysis and quantification of participants’ in-session statements [33]. Detailed and clear descriptions of the data analysis procedures were presented in two studies [31, 32], whereas in one study the description was unclear [33] and in another study minimal details were provided about the data analysis [20].

In most studies, some context and setting were described. The recruitment procedure was reported in all studies, but information on sample demographics was relatively sparse across studies, with one study lacking even the a minimal necessary information such as participating mothers’ age and marital status [33]. Informed consent from participants was only reported in two studies [31, 33] and the issue of protecting participants’ confidentiality and anonymity was mentioned in one study alone [32]. Only one study mentioned the potential influence of the therapists or researchers on the findings, “since the patients who were interviewed were selected in consultation with group facilitators” [31]. There were no other indications of the researchers’ reflexivity regarding their potential influence on the findings in terms of their own bias, perspective, role, and interactions with participants.

Finally, in all studies sufficient data were presented to support the findings, ranging from relatively long quotations [31] to short quotations [20, 32], and paraphrases of the participants’ statements [33]. A procedure to enhance the credibility of the findings was only reported in one study [32], where the themes were “confirmed in consultation with a psychodrama expert.” It is not clear, however, if this expert was blind to the study aims.

Mixed methods studies

As shown in Table 2, of the seven mixed methods studies, four had a single group design [23, 34–36]. Two studies had nonrandomized group comparison designs: one with a treatment-as-usual control group matched for age, gender and diagnosis to minimize selection bias [37], and one with an unmatched and unspecified control group [38]. One mixed methods study had a randomized group comparison design and a 4-week control group on “effective learning and efficient study methods” which was shorter than the 12-week PD group [39].

The number of participants in the mixed methods studies ranged from six to 76 (M = 25, SD = 24). Most of the studies involved adult participants. One study involved adult inmates in the UK who were approaching their release date [36], one study involved adults in Turkey with HIV or AIDS [23], one study involved adults in Australia with unresolved painful emotional experiences [34], and one study involved women in Turkey with a diagnosis of infertility [38]. One study involved college students in South Korea struggling with life adjustment and interpersonal relationships [35], one involved Master’s students in Turkey with interpersonal difficulties [39], and one study involved adolescents in Italy with psychiatric disorders [37]. The setting of two studies was universities [35, 39], one study was conducted at a university care center [23], one was conducted at a fertilization unit of a university hospital [38], one in prisons [36], one in private practice [34], and one was conducted in a semi-residential public service setting [37].

In all seven studies PD treatment was delivered in a group format, of which one study used a semi-structured analytic PD group [37] and one used a protocol for a group PD combined with cognitive-behavioral techniques [36]. All the other studies generally followed the classical PD approach with a standard structure of three PD phases. In three studies, session duration ranged from 1.25 to 3 hours per week, and the number of sessions ranged from 8 to 12 [37–39]. One study delivered six weekly hours for 10 weeks [23]. Two studies were delivered in a concentrated format of a 2- or 3-day intervention [36] and a 2.5-day workshop [34]. Two of the seven mixed methods studies did not provide information concerning the qualifications of the therapists [35, 36]. Only one study indicated that adherence to PD treatment was monitored for treatment fidelity [34] and another study indicated that the PD therapist received clinical supervision during treatment implementation [23].

Qualitative data were collected using post-treatment individual interviews and/or feedback forms in three studies [34–36]. Two studies included direct observation during the sessions in addition to interviews [38, 39], one had focus group interviews both before and after treatment [23], and one presented case study descriptions of clients [37]. Detailed and clear descriptions of qualitative data analysis procedures were absent in all studies, and only one study specified the procedure used for qualitative data analysis as “descriptive analysis” [23]. Quantitative data were collected in all seven studies before and after treatment (pretest-posttest), and one study also collected data at 2-week and 3-month follow-ups [34]. Regarding quantitative data analysis, because of the small sample sizes and/or non-normally distributed variables, four studies computed non-parametric tests: a Mann-Whitney U test for independent samples and/or a Wilcoxon matched-pairs signed-rank test for paired samples [23, 37–39]. Two studies computed t-tests [35, 36], of which one also computed a repeated measures analysis of variance [36], and one study computed linear mixed models analysis but without a detailed description of the analysis [34]. Only two studies reported effect sizes. The study with adult inmates reported the highest (medium) effects for improved self-efficacy (d = 0.67) and social and friendship skills (d = 0.54) [36]. The study on adults with HIV or AIDS reported a large effect size on mental health improvement (d = —.82) [23].

In all studies, some context and setting were described but more details could have enhanced the findings' transferability. The recruitment procedure was reported in all studies and information on the sample demographics varied across studies but was satisfactory. Obtaining informed consent from participants was indicated in five studies, with the exception of two [39] of which one only stated that participation was voluntary [35]. The issue of protecting participants’ confidentiality and anonymity was mentioned in one study alone [34] but two other studies referred to the group as a “confidential environment” [23] or to the “confidentiality of the sessions” [35].

Three studies mentioned the potential influence of the therapists or the researchers on the qualitative findings; e.g., through their interactions with clients/ interviewees [23, 34, 36], but only one implemented a procedure: “To reduce potential bias arising from dual roles, protagonists who had been directed by the first author were interviewed by other team members” [34]. In all studies sufficient qualitative data were presented to support the findings, ranging from relatively long quotations [34–36], to short quotations [23, 37, 38], or a combination of both [39]. Procedures to enhance the credibility of the qualitative findings were not reported in any of the studies. Finally, while only two studies were defined as “mixed methods” [38, 39], the collection of both qualitative and quantitative data was appropriate for the purpose of all the studies. In all studies, the integration of qualitative and quantitative data was presented in the Discussion section to a varying degree. No study included an integrated mixed method question or hypothesis.

Quantitative single group studies

As seen in Table 3, the number of participants in the six single group studies ranged from 12 to 30 (M = 21, SD = 7.76). Three studies involved children and/or adolescents [29, 40, 41] and two studies involved university students: one on anger management problems in Turkey [42] and one on mental health conditions in Italy [43]. One study consisted of both adults with mental health conditions and students in Israel [44]. The setting of two studies was schools [29, 41], three studies were conducted at universities [42–44], and one study was conducted at an outpatient clinic [40].

In five studies, PD treatment was delivered in a group format, of which one study used a 3-stage protocol for psychoeducation sessions based on PD and some cognitive-behavioral techniques [42], in one group analytical PD was used with a protocol that was not reported [43], and the other four group PD studies followed the standard three PD phase structure. One study was delivered in an individual format, but some sessions were also attended by the adolescent’s or child’s parents [40]. In three studies, session duration ranged from 1.25 to 2 hours per week, and the number of sessions ranged from 16 to 40 [29, 43, 44]. In two studies PD was delivered twice a month [40, 42] and in one study in a 3-day format with unspecified hours [41]. Information concerning the qualification of the therapists was provided in only three of the six studies [29, 43, 44]. One study indicated that the PD director was “one of the researchers” [41] and another study indicated that participants were treated “by the author” [40]. Only one study described a procedure for ensuring implementation fidelity [44] and one study indicated that the PD therapist received clinical supervision during treatment implementation [29].

Data were collected in four studies before and after treatment [40– 43], and one of these studies also collected data at a 22-month follow-up [40]. Two studies had a single case design (also termed “single subject design”) with repeated measures over the course of the treatment, of which one also had three post-treatment measures [44] and one measured in-session changes [29].

Data analysis included a non-parametric Mann-Whitney U test for independent samples and/or a Wilcoxon matched-pairs signed-rank test for paired samples [40, 42]. Analysis of variance and paired t-tests were computed in two studies [41, 43]. One study with a single case design computed a hierarchical linear model that took the nested data structure into account and compared aggregated measures at three time points [44]. Another study with a single case design computed a hierarchical linear model and odds ratio, which is an effect size statistic reflecting the odds of clients’ dramatic engagement in predicting their productive in-session behaviors, with effects ranging from small (0.70) to medium (4.30) [29]. No other studies reported effect sizes.

The recruitment procedure was reported in all the studies. Information on sample demographics varied across studies but was generally satisfactory, with the exception of two studies that only reported age and gender [40, 41]. In three of the six studies it was indicated that participants provided informed consent [40, 41, 43].

Non-randomized group comparison

The number of participants in the two non-randomized group comparison studies ranged from 23 to 40 (M = 32, SD = 12), as seen in Table 4.

One study involved women coping with the first episode of major depression was conducted at a university polyclinic in Turkey [19]. This study included three groups and compared the effects of group PD with anti-depressant medication to the effects of a group with antidepressant medication alone and to a healthy control group of volunteers with similar ages and education levels. The second study involved adolescents at risk that was conducted at a middle school in Israel [45]. This study included two groups and compared the effects of group PD to a waiting list control group.

Regarding the treatment, the study at the university polyclinic had a loosely structured treatment with a session duration of three hours per week, for 16 sessions [19]. The study at the school did not follow a treatment protocol, and session duration was 1.5 hours per week for 16 to 20 sessions [45]. In both studies, information concerning the qualification of the therapists was provided and both studies indicated that the therapist was given clinical supervision during treatment implementation.

Data were collected in both studies before and after treatment (i.e., pretest-posttest). In one study, due to small sample size and non-normally distributed variables, the data analysis included a non-parametric Kruskal Wallis test for the comparisons of the three groups and a Wilcoxon matched-pairs signed-rank test for paired samples [19].The second study included a mixed‐design repeated measures analysis of variance and process-outcome analyses based on correlations between the change scores for the process and outcome variables [45]. Only the latter study reported effect sizes, with effects ranging from medium (.11) to large (.28) partial eta-squared. In both studies, the recruitment procedure was reported but information on sample demographics was more detailed in one study [45] than the other [19]. In both studies participants provided informed consent.

Randomized group comparison

As seen in Table 4, the number of participants in the 11 randomized group comparison studies ranged from 16 to 113 (M = 53, SD = 30). Four studies were on adolescents, of which three were on adolescents with aggression problems in Turkey (reported as distinct studies [46–48]) and one on Latino immigrant adolescents in the US (the same study was reported in two separate articles [21, 22]). The remaining seven studies were on adult smokers in Turkey [49], men with opiate dependence in Iran [50], adults with burnout who work in public services in Finland [18], oncology nurses coping with burnout in Turkey [51], adults with Parkinson’s disease in Brazil [52], and adults with panic disorders in Bulgaria [30]. One study was on university students with low subjective well-being and high hopelessness in Turkey [53].

The settings of three studies were schools [46–48], three studies were conducted at health service centers [18, 30, 50], two studies at hospital clinics [51, 52], and two at universities [49, 53]. One study was conducted in community settings [21, 22].

In all studies PD treatment was delivered in a group format, of which five followed the standard three PD phase structure [46–48, 50, 53]. Two treatments were semi-structured [49, 51], one treatment had a structured protocol [21, 22], whereas in three studies the content of the PD treatment was unclear [18, 30, 52].

In most studies (n = 8 of 11, 72%), session duration ranged from 1.5 to 3 hours per week, and the number of sessions ranged from eight to 25 [21, 22, 30, 46–49, 51, 53]. One study delivered a 2-hour group PD for 12 sessions over six weeks [50], another delivered 6-hour group PD sessions over 17 days [18], and in one study a 1.5-hour group PD was delivered twice a month for 12 sessions [52].

Therapists’ qualifications were reported in eight studies, of which only six reported that the PD groups were led by psychodramatists [22, 30, 46, 49, 51, 53] and two studies reported that the PD groups were led by a psychologist and a physiotherapist [18], and a psychologist [52]. The remaining three studies did not report any information about the therapists’ qualifications [47, 48, 50]. Only two studies indicated the PD therapist received clinical supervision during treatment implementation [22, 46].

Data collection in most (9 of 11, 81%) of the randomized group comparison studies included before and after treatment measures and a follow-up measure that ranged from 10 weeks to 1 year after the treatment, with the exception of two studies that only included pretest-post measures [50, 52]. Data analysis in most studies (7 of 11, 64%) used parametric tests such as analysis of variance and/or t-tests, [18, 30, 46, 47, 50–52], as well as regression analysis [21, 22]. Three studies used non-parametric tests, including a Mann-Whitney U test for independent samples, a Wilcoxon matched-pairs signed-rank test for paired samples, and/or a Kruskal-Wallis test for independent samples, a chi-square test or a Fisher exact test for categorical data [48, 49, 53]. Three studies reported moderate to large effect sizes [18, 21, 22, 50], and one study reported an odds ratio indicating a small effect size [49].

The recruitment procedure was reported in all studies and information on sample demographics varied across studies but was generally satisfactory, with the exception of five studies that only reported age (or grade), gender, and/or scores on a screening instrument [18, 46–48, 53]. In two studies it was indicated that participants provided informed consent [51, 52] and two studies indicated that participants volunteered [46, 47, 53].

Risk of bias assessment

This section reports the risk of bias analysis results for 12 randomized group comparison studies (11 in Table 4, and the Dogan et al. 2010 study in Table 2). A summary of the risk of bias analysis is presented in Table 5 and the proportion of studies for each of the six potential sources of bias is presented in Fig 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Proportion of studies for each of the six potential sources of bias.

https://doi.org/10.1371/journal.pone.0212575.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Risk of bias assessment results per study.

https://doi.org/10.1371/journal.pone.0212575.t005

As can be seen, none of the studies was identified as having a low risk of bias across all six sources of bias. Evidence for potential selection biases (both “random sequence generation” and “concealment of allocation to conditions”) were unclear in the majority of studies (75%, 9 of 12). One study was at low risk of these two selection biases because it had a restricted block randomization method and used sealed envelopes “prepared before enrollment by the head nurse” [49], suggesting that participants and investigators enrolling participants were blind to the allocation to conditions. A high risk of these two sources of selection biases was evident in a study where clients were randomized “into the two different intervention groups but not into the control group” [18]. Thus, randomization was incomplete and clients or investigators could have possibly foreseen the allocation to conditions. Another study was at high risk for inadequate allocation concealment because it was explicitly defined as “an open” clinical study in which the allocation to conditions was not concealed [30].

The blinding of participants and personnel (i.e., clients and therapists) is intended to prevent the direct and indirect influence of their expectations or motivations on the study outcomes. However, this type of performance bias is impossible to avoid in psychotherapy: therapists know what treatment they are delivering, and clients not only are aware of receiving that treatment but also have expectations concerning the treatment [27]. For this reason, psychotherapy reviews have considered there is a low risk of bias on the basis of two adjusted criteria: (a) the use of an active control group with a credible treatment involving comparable treatment expectancies, and (b) the assessment of client-perceived “treatment expectancies or credibility prior to or early (i.e., after the first session) in treatment and in which the treatment group had equal (zero difference) or lower expectancies or credibility than the control group received” (see Appendix C, p. 4, in [28]). In the present review, none of the studies was identified as having a low risk of performance bias according to these criteria. Evidence for potential blinding performance bias was unclear in the majority of studies (67%, 8 of 12) that had an active control group but did not measure client-perceived treatment expectancies or credibility. Four studies (33%) were at high risk of performance bias because they had an inactive control group and no measures of treatment expectancies or credibility [18, 50–52].

Attrition bias refers to the mishandling of incomplete outcome data. There was a low risk of attrition bias in five studies (42%) with no reported dropouts, appropriate imputation of missing data, and/ or unlikeliness that the missing data were related to the outcome [30, 48, 50, 52, 53]. The risk of attrition bias was unclear in four studies (33%) due to insufficient reporting of attrition [46, 47, 49, 51]. There was a high risk of attrition bias in three studies that had a dropout rate exceeding 20% and no imputation of missing data [18, 21, 22, 39].

Selective outcome reporting bias refers to cases where outcomes that were collected according to the methods section (or to a pre-study published protocol or a clinical trial’s registry) are not reported. There was a low risk of reporting bias in only one study [39], and an unclear risk in five studies [21, 22, 30, 46, 47, 49, 50]. There was a high risk of reporting bias in five studies where not all of the pre-specified primary outcomes or not all of the results related to the outcome were reported [18, 48, 51–53].

Treatment implementation is a type of performance bias that refers to deviations from intended treatment that are likely to have affected the outcome and to the qualifications of the therapist delivering the treatment [27]. There was a low risk of treatment implementation bias in six studies because both the psychodramatists’ qualifications and the treatment content were reported [22, 39, 46, 49, 51, 53]. The remaining six studies were at unclear risk of treatment implementation bias because the qualifications of the therapist were not reported [47, 48, 50], the therapist was not identified as a psychodramatist [18, 52], and/or a psychodramatist delivered the therapy but no content was described [30]. None of the studies was identified as having a high risk of treatment implementation bias.

Qualitative studies

The four qualitative studies presented sufficient excerpts from the data which showed that overall, PD can contribute to clients’ self-worth and promote self-awareness, self-expression, and better perceived coping with difficulties. The descriptions of participants’ characteristics, the contexts, and settings could have been more detailed to allow for comparison with other contexts and settings and to enhance the transferability of the findings. According to the American Psychological Association’s new Journal Article Reporting Standards (“APA JARS”) for Qualitative Research, “transferability of findings in qualitative research to other contexts is based on developing deep and contextualized understandings that can be applied by readers” [60]. The rationale for choosing a specific data analysis method varied across studies and, generally, descriptions could have been clearer and more detailed. In addition, the researchers did not provide a reflexive self-description of their views and backgrounds (e.g., demographic/cultural characteristics, values, experience with the phenomena, training, etc.), or whether and how reflexivity was used to mitigate their potential influences on the participants and/or the interpretation of the findings. It is worth noting that one study applied a procedure to enhance the credibility of the findings by confirming the emergent themes with a psychodrama expert [32]. Future studies could enhance the credibility of the findings by implementing procedures such as data triangulation, disinterested peer debriefing, external auditing, or member checking (also called “informant feedback” or “respondent validation”).

Mixed methods studies

Mixing qualitative and quantitative methods strengthens the internal and external validity of the findings, in that the advantages of each method complement one another, and their shortcomings are considerably offset [62]. In the seven mixed methods studies, data integration was presented in the Discussion section, which was useful for interpreting the findings based on the two datasets as well as for giving a (qualitative) voice to participants and ensuring that the quantitative results reflected their experiences. An “integrated” mixed method question or hypothesis which directly addresses the mixing of the quantitative and qualitative strands of the study were not put forward in any of the studies. An example of an integrated question that conveys the methods of the study is “Do the qualitative data help explain the results from the initial quantitative phase of the study?” and an example that conveys the content of the study is “Do clients identify helpful aspects in therapy that shed light on why or how therapy was effective?” In addition, most studies did not contain a statement about the rationale for choosing a mixed methods design such as a more comprehensive and nuanced understanding of the research problems. Researchers should also be encouraged to state the type of mixed methods design used, and specify the timing of the data collection (concurrent or sequential) as well as the emphasis (equal or unequal) on each data type. Examples of mixed method designs include Convergent Parallel design, Exploratory Sequential design, and Explanatory Sequential design [63]. One mixed methods approach that has been adopted by some PD researchers in Europe [64] is the Hermeneutic Single-Case Efficacy Design which integrates quantitative outcome and weekly change data with clients’ qualitative accounts of change over therapy, while examining alternative non-therapy explanations for changes in therapy [65].

Quantitative studies