Introduction

Clinical supervision is defined as “the formal provision of an intensive relationship-based education and training by qualified health practitioners that is case-focused and which supports, directs, and guides the work of supervisees” (^{Milne, 2007}, p. 440). Training is a more general concept that refers to the acquisition of work-related competencies, and clinical supervision focuses on the therapists' work with their own cases and the provision of supervision guidelines (^{Milne, 2007}). Training, and in particular clinical supervision, for managing the techniques and the forging of an adequate therapeutic relationship has been highlighted as one of the factors that can explain success of psychological therapy (^{Erekson et al., 2017}; ^{Snowdon et al., 2017}). Several monographs have been published, including the Wiley International Handbook of Clinical Supervision (^{Watkins & Milne, 2014}), which provides a comprehensive overview of the research and clinical work up to date in this area. In parallel to the previous publication, the American Psychological Association (APA) has released a series of guidelines for the adequate practice of clinical supervision (^{APA, 2014}). These publications highlight the growing commitment to the extension of clinical supervision to therapists and the best ways to do so.

Different methods of supervision have been documented in the literature from multiple theoretical orientations. From the cognitive-behavioral approach (CBT) discussion, role play and role play modeling are among the most widely implemented. Discussion is the most widespread method of supervision regardless of its usefulness, as it seems to be preferred for its ease of application but shows lower compliance with the guidelines (^{Bearman et al., 2013}). Research on variables associated with compliance with the guidelines by the supervisees in therapeutic sessions is scarce. Confidence with the supervisors is among the most described variables that could influence the perceived impact of the guidelines (^{Bernard & Goodyear, 2014}). Research on the aforementioned variables is focused on the relationship between supervisors and supervisees, in many cases relying on simple measures such as a single item (^{Kavanagh et al., 2003}).

Given this scenario, this is an emerging area of research, still with important work ahead. Basic gaps are still to be filled in the literature in order to understand clinical supervision and its usefulness, such as finding out whether therapists actually comply with the guidelines they receive and the mechanisms necessary to favor it (^{Bearman et al., 2013}). Understanding this phenomenon will lead to testing, in future approaches, other effects of clinical supervision, such as favoring the success of psychological therapy (^{Callahan et al., 2009}). For this purpose, we will use observational methodology. Its advantages have been pointed out in the behavioural sciences (^{Pérez, 1996}). Despite this, its use is not widespread due to its implementation difficulties, such as greater time consumption for observer training and data recording, and the economic costs of the acquisition of coding software (^{Heyman et al., 2014}). Natural observation, correctly used, entails the advantage of maintaining adequate external validity (^{Bakeman & Quera, 2011}). Using observational methodology, we can approach to explore the compliance with the supervision guidelines avoiding falling into the limitations of other studies, in which the experimental manipulation has yielded results alien to the clinical reality (^{Chambless & Ollendick, 2001}).

The objectives of this study are: (1) to contribute to the knowledge of clinical supervision using observational methodology, exploring the compliance with the guidelines of novice therapists, as well as variables that could contribute to this; and (2) to explore the characteristics of clinical supervision at the psychology clinic of the Universidad Autónoma de Madrid (CPA-UAM). Observational designs with similar objectives have been carried out with N = 1 designs (e.g., ^{Milne et al., 2003}), but our approach is innovative as it considers the supervision guidelines as the unit of analysis, allowing a more sophisticated treatment of the data. According to the aims described, we developed several hypotheses, as follows:

Hypothesis 1: The most frequently used supervision method is discussion. Given the predominance of this method, as reported in the scientific literature (e.g., ^{Bearman et al., 2013}), it is expected that a greater number of supervision guidelines will be provided by discussion compared to role play and role play modeling, as assessed by a self-developed rating scale.

Hypothesis 2: The higher the therapists' confidence with the technical supervision provided by the supervisors, the greater their compliance with the guidelines. It is expected that guidelines categorized with a rating scale as “compliance” correspond to higher scores in confidence with supervision on a self-report.

Hypothesis 3: The higher the therapists' rating of the usefulness of supervision methods, the greater their compliance with the guidelines. It is expected that guidelines categorized with a rating scale as “compliance” correspond to higher scores in confidence with the supervision methods on a self-report.

Hypothesis 4: The higher the therapists' confidence to ask questions to their supervisors, the greater their compliance with the guidelines. It is expected that guidelines categorized with a rating scale as “compliance” correspond to higher scores in confidence to ask questions on a self-report.

Method

Sample

Six novice CBT therapists (supervisees), in their first year of residency (R1) at the CPA-UAM and seven supervisors, teachers of the Faculty of Psychology or therapists at the CPA-UAM with more clinical experience than the R1s, participated. Recordings of forty supervision sessions and 80 therapy sessions (the two sessions that followed each supervision session) were collected, which ranged in length from 12 min to 1 hour and 11 min (M = 35.4 min); 31 clients of the CPA-UAM, both adults and children, with different diagnoses participated in the study. The training program consisted of a three-year residency. Between four and six novice therapists enter the first year of residency (R1), and only two of them continue as second (R2), and third-year (R3) residents. Residents participate in periodic supervision sessions, with R1s receiving 8 hours per week. Table 1 includes some characteristics of the supervisors and supervisees.

Table 1. Supervisees and Supervisors Sociodemographic Data 

Note. T = therapist; S = supervisor.

Supervisees, supervisors, and clients signed informed consent specific to the current research project in which they authorized the researchers to collect, on video tape, the sessions in which they participated. Data collected were encrypted by reversible dissociation through a code, known only by the researchers, thus guaranteeing confidentiality. The research project in which this study is included was approved by the ethics committee of the Autonomous University of Madrid.

Results

A total of 603 supervision guidelines were collected. The results are presented below, organized according to the hypotheses described.

Hypothesis 2: The Higher the Therapists' Confidence with the Technical Supervision Provided by Supervisors, the Greater the Compliance with Guidelines

Homoscedasticity could not be assumed, F(3, 366) = 12.057 (p < .001), so we used Welch's F, a robust alternative of the F-statistic for samples with different group sizes. A significant difference was found between at least two compliance groups F(3, 49.324) = 6.304 (p = .001, est. ω² = .042), with a small effect size. Post hoc comparisons, performed using the Games-Howell test, showed that the guidelines rated “total compliance” received a significantly higher score in confidence with the supervision provided by supervisors than the ones rated “non-compliance”, the mean difference being 0.512 (p < .001, CI 95% [0.20, 0.82]). Table 2 shows the confidence means in the compliance groups, the pairwise comparisons and the significance of their differences.

Table 2. Games-Howell Test Pairwise Comparisons: Confidence with Technical Supervision 

Note. TC = total compliance; PC = partial compliance; NC = non-compliance; FI = future implementation.* p < .05.

Hypothesis 3: The Higher the Therapists' Rating of Usefulness of Supervision Methods, the Greater the Compliance with Guidelines

Homoscedasticity could not be assumed, F(3, 362) = 3.288 (p = .021). A significant difference was found between at least two compliance groups F(3, 49.763) = 2.977 (p = .04, est. ω² = .016), with a small effect size. Post hoc comparisons, performed using the Games-Howell test, showed that the guidelines rated “partial compliance” received a significantly higher score in confidence with the supervision provided by supervisors than guidelines rated “non-compliance”, the mean difference being .57 (p < .036, CI 95% [0.26, 1.11). Table 3 shows the confidence means in the compliance groups, the pairwise comparisons and the significance of their differences.

Table 3. Games-Howell Test Pairwise Comparisons: Confidence with the Methods of Supervision 

Note. TC = total compliance; PC = partial compliance; NC = non-compliance; FI = future implementation.* p < .05.

Differences in sample sizes between the compliance groups in which a significant difference was found may have led to biased results (^{Delacre et al., 2019}), so we grouped compliance groups in: (1) compliance (total compliance and partial compliance) and (2) non-compliance (non-compliance and future implementation). Guidelines rated “compliance” received a significantly higher score in confidence with the supervision methods (n = 176, M = 8.119, SD = 1.076) than “non-compliance” (n = 190, M = 7.842, SD = 1.320), t(359.273) = 2.203 (p = 0.028, η² = .01), with a small effect size.

Hypothesis 4: The Higher the Therapists' Confidence to Ask Questions to Their Supervisors, the Greater the Compliance with Guidelines

Homoscedasticity could not be assumed, F(3, 362) = 3.121 (p = .026). There is no evidence of differences in confidence to ask questions between the compliance groups F(3, 51.037) = 0.580 (p = .631, est. ω² = -.003). Mean confidence to ask questions scores for each compliance group were: (1) total compliance (n = 145, M = 8.69, SD = 0.829), (2) partial compliance (n = 32, M = 8.72, SD = 0.851), (3) non-compliance (n = 175, M = 8.80, SD = 0.945), and (4) future implementation (n = 14, M = 8.86, SD = 0.663).

Discussion

In light of the results, the predominant method of supervision at the CPA-UAM is discussion, and there are higher ratings of confidence with technical supervision and usefulness of the methods in the supervision guidelines complied by the supervisees. No relations were found between a more relationship-related confidence type (confidence to ask questions) and compliance with the guidelines. The results will be discussed below.

As we predicted in hypothesis 1, the most used supervision method was discussion, a phenomenon reported in the scientific literature, even though other methods, such as role play and role play modeling, are more advisable (^{Bearman et al., 2013}). This could be due to the ease of implementation of this method and time constraints that may exist in psychology clinics. Eight hours of supervision are carried out weekly at CPA-UAM with the six R1 therapists in which they supervise their cases. The predominance of discussion may be due to the need to supervise several cases in a short period of time: based on this data, a reorganization of the schedule could be advisable, accompanied by training for the supervisors on the benefits of implementing other methods.

Hypotheses 2, 3, and 4 were proposed to explore the relations between compliance with the guidelines and some variables of supervisees' confidence with supervision, a phenomenon loosely analyzed in the literature. In general terms, we found that guidelines assessed as “compliance” had higher scores in terms of confidence with technical supervision (hypothesis 2) and perceived usefulness of the methods (hypothesis 3), both technical aspects of supervision. In comparison, a more relationship-related confidence type (confidence to ask questions, hypothesis 4), was not related to compliance with the guidelines. The scientific literature is unclear as to what confidence with supervision entails, so the differences found can guide the approach of future studies. In studies planned by our research team, a coding system is being used to analyze the verbal interactions between supervisors and supervisees, allowing us to analyze aspects such as the supervisors' styles of supervision. This will help us understand the results found to a greater extent.

The main contribution of this study is the proposal of an innovative approach to the study of the compliance with the supervision guidelines using observational methodology. The guidelines were used as the unit of analysis as opposed to previous observational approaches, that used the therapists as the unit of analysis with the proposal of N = 1 designs (^{Milne et al., 2003}). It has been possible, therefore, to perform more sophisticated analyses, helping the generalizability of the results, a limitation of some N = 1 designs. Moreover, this study has allowed for an in-depth study of the training method of the CPA-UAM. The final aim of this research line is the elaboration of a guide for the training of supervisors, given that nowadays they face this task guided only by their own clinical experience. Future approaches could focus on the exploration of the level of compliance between therapists and supervisors with different levels of experience.