Effectiveness of Outpatient Rational Emotive Behavior Therapy Over One Decade

Randomized controlled trials (RCTs) are the gold standard in testing the efficacy of any psychological treatment (i.e., the treatment works in controlled conditions and maximizes internal validity). To date, CBT (REBT alone or considered as a part of CBT) has been among the psychological interventions empirically investigated most frequently and has received the most empirical support (7, 8). The efficacy of CBT has been evaluated in several meta-analyses (9–13), with positive results. More recently, the evidence supporting CBT has been strengthened by several rigorous clinical trials (14, 15) and meta-analyses (16–18).

However, questions remain concerning the procedures used for increasing internal validity in RCTs—procedures that may compromise the external validity of the results. Specifically, these questions pertain to generalizability across patients, clinicians, and treatments used and about whether these characteristics of clinical trials are representative of clinical practice. Moreover, some studies (19, 20) have found that up to 70% of patients who could benefit from depression treatment would not have been eligible to participate in a RCT efficacy study. The most common reasons for excluding patients from RCTs have been that they received subclinical diagnoses, had comorbid psychiatric diagnoses, or received adjunct treatments, such as medication. Because of these exclusionary problems and because efficacy studies fail to provide information about the transportability of interventions to more ecologically valid conditions, psychotherapy research requires an emphasis on effectiveness trials that evaluate interventions implemented under naturalistic conditions (21, 22).

Effectiveness research investigates whether interventions that were tested in RCTs attain similar benefits in naturalistic settings; such research can include pretest-posttest, quasi-experimental, or experimental designs. According to Stewart and Chambless (23), a study design can be categorized as having external validity when it achieves at least one of the following characteristics: it captures data in clinically representative settings (e.g., outpatient clinics, mental health centers), uses clinically representative therapists (e.g., practicing clinicians who do this job daily), or studies clinically representative patients (e.g., has few exclusion criteria). Additionally, the transdiagnostic approach to the treatment of clinical problems reflects the current at-hand orientation in real practice, as opposed to the model of selected interventions for specific disorders that characterizes randomized trials.

Several reviews have synthesized the results of CBT effectiveness studies. Stewart and Chambless (23) conducted a meta-analysis of 56 CBT effectiveness studies for adult anxiety disorders to investigate whether CBT tested under well-controlled conditions can be generalized to less controlled settings. The effect sizes from pretest to posttest for disorder-specific symptom measures were large, suggesting that CBT for adult anxiety disorders is effective in clinically representative conditions. The findings demonstrated that results from effectiveness studies are similar to those obtained in efficacy trials. Hans and Hiller (24) conducted a meta-analysis of 34 CBT effectiveness studies of outpatients receiving individual and group therapy for unipolar depressive disorder. Results showed large posttreatment effect sizes for reducing depression severity with outpatient CBT and moderate-to-large effect sizes for secondary outcomes. In a study of general psychological distress, Twomey et al. (25) demonstrated the effectiveness of a computerized CBT program (MoodGYM, d=0.48) in reducing psychological distress among a sample of 149 public mental health service users. Hickie et al. (26) found effect sizes in the same range (d=0.4) when measuring the level of psychological distress from pre- to posttest. Although the results were consistent with previous findings showing that individual and group outpatient CBT can be effectively transported to clinical practice, the study also found that patients encountered in clinical practice showed less improvement in depressive symptoms than did participants in RCTs.

A commonly used method in ecological studies of psychotherapy, especially with center-based samples, is to track symptom change during therapy. The Outcome Questionnaire–45 (OQ-45) (27) has become one of the 10 instruments most frequently used by practitioners to measure clinical outcomes (28). Research using this type of measure has helped researchers identify whether patients receiving high doses of therapy would be no more likely to experience a clinically significant change than those who receive low doses (29–30). Baldwin et al. (31) evaluated the dose-effect relationship and model of change by examining the relationship between the rate of change and total dosage of sessions among 4,676 patients who received individual psychotherapy. They showed that 1,242 patients achieved clinically significant change (measured with the OQ-45) and that faster rates of change were registered by the patients receiving small doses of treatment, whereas slower rates of change were linked to receiving larger doses of treatment.

Other important variables that can be investigated in such naturalistic studies of psychotherapy effectiveness are dropout rates and course of patients’ treatment response. High dropout rates (24%–66%) have been reported by studies (32, 33) conducted in mental health centers. Low motivation for change was identified as a primary reason for dropout (34). Studies addressing the long-term outcomes of psychotherapy found that early responders (e.g., patients improving within the first three sessions) had better outcomes at therapy termination and maintained their therapy gains (35). Moreover, an early positive response to psychotherapy has been found (36) to be a powerful predictor of outcomes among various samples. Also, some studies (37–39) have documented the effectiveness of brief time-limited therapy (eight to 16 sessions or five sessions of CBT) (30, 37). It has been suggested that about 50% of treated patients in routine care might recover if they received about 18 to 21 sessions of psychotherapy (40), although about 50% of clients will show reliable improvement after only seven sessions. However, it is also known that deterioration rates for patients in routine practice settings can be as high as 14% (41), whereas the response rate (recovered and improved patients) for patients with clinical symptoms participating in 20 sessions of therapy is around 40% (31).

To our knowledge, systematic studies to date have not yet investigated the effectiveness of REBT. This research gap is surprising, because REBT has been practiced as the original form of CBT, and, from the beginning, the REBT literature has more closely resembled an effectiveness framework than an efficacy framework. Indeed, the Albert Ellis Institute (AEI) has for many decades offered services to the community that are based on the REBT model, focus on clinical problems as they appear in real life, are provided by therapists whose main job is conducting therapy, and implement both the transdiagnostic and categorical diagnostic approaches. Thus, the AEI seemed to be the perfect setting to evaluate the effectiveness of REBT with the purpose of understanding the impact of CBT more generally. Therefore, in this study, our main hypothesis was that the use of REBT can improve psychological functioning, expressed as lower symptom severity, of a sample of outpatients in naturalistic conditions.

Data from several hundred adult patients of the AEI who sought REBT from 2007 to 2016 were culled from AIE records for this study (child patients will be analyzed in a future study). The yearly distribution of patients included in the study, by year of intake, is presented in Figure 1.

We limited our analyses to patients who received at least one therapy session; we considered 20 sessions of therapy as posttest (a prototype number for the CBT sessions). Thus, we included 349 patients who had participated in at least one session of therapy after intake and had provided the data necessary for our analyses. Of the participants, 55% (N=192) were women, and 44% (N=154) were men (<1% [N=3] missing data); ages ranged from 18 to 75 years (mean±SD=35.39±11.44). Of the participants, 83% (N=290) completed at least three sessions of therapy, and 17% (N=59) completed all 20 sessions. The mean intake OQ-45 score among participants was 74.05±23.39, which has been previously considered typical for outpatient therapy settings (27).

Therapists who participated in the study were either staff therapists at the AEI or trainees. All staff therapists had received at least 2 years of training at the AEI. The trainees were psychology doctoral students completing an externship, internship, or postdoctoral training. All therapists completed 3 days of primary training and 4 days of advanced training in REBT before they began treating study patients. Training required the therapists to read seminal books in REBT; attend eight lectures on REBT theory, techniques, and applications to clinical problems; and complete eight peer-counseling sessions that were observed by a supervisor who provided in vivo feedback. The therapists received weekly supervision, which included reviewing recorded sessions and receiving feedback on their therapy sessions, from an experienced licensed psychologist as well as an REBT- and CBT-certified supervisor. The licensed staff therapists attended clinical case conference meetings once a month to discuss their cases, and unlicensed staff had weekly supervision by a licensed psychologist and a certified REBT and CBT supervisor.

We used the OQ-45 (27) to measure patient functioning. The OQ-45 is a widely used instrument for examining the effectiveness of psychotherapy over time, and it is considered useful for tracking changes in patients’ symptoms on a session-by-session basis (42). The OQ-45 consists of 45 items each scored on a 5-point Likert-type scale. The instrument can be scored as a total across three subscales assessing symptom distress, social role functioning, and interpersonal relationships. The symptom distress subscale measures general emotional and behavioral symptoms, such as depression, anxiety, stress, substance abuse, and suicidality; scores range from 0 to 88, with higher scores indicating higher levels of distress. The social role functioning subscale measures patients’ work relations and leisure activities; scores range from 0 to 36, with higher scores indicating higher dissatisfaction in social roles. The interpersonal relationships subscale measures patients’ satisfaction with interpersonal relationships, such as marital and family relationships and friendships; scores range from 0 to 44, with higher scores indicating greater problems in interpersonal relationships. The total OQ-45 score, ranging from 0 to 180, has been found to be reliable and valid (43) and distinguishes between individuals with clinical and nonclinical symptoms.

Patients were enrolled between 2007 and 2016 at the AEI in New York City. The only inclusion criterion was being older than age 18. We based this decision on the fact that the transdiagnostic approach is dominant in real clinical practice as opposed to in randomized trials (44). Patients were assigned to a therapist on the basis of the availability of time slots that matched the patients’ and therapists’ schedules. The study was approved by the institutional review board of the AEI.

Patients received an intake packet of questionnaires prior to their first session, which they began completing before their session and finished after that session or a subsequent session. The questionnaires included the Millon Clinical Multiaxial Inventory–III (45), the Psychiatric Diagnostic Screening Questionnaire (46), the Attitudes and Beliefs Scale–2 (47) (a measure of irrational and rational beliefs), and a four-page case history background questionnaire created by AEI staff (48). The therapists used this information to develop a case conceptualization for each patient. Patients completed the OQ-45 before entering therapy and then before each session to monitor their progress.

All patients received REBT at the AEI. Each session lasted 50 minutes, opened with a “mood check,” and continued with a review of the previous session and homework follow-up. The therapist and patient set the agenda and approached it based on the transdiagnostic ABC(DE) model (4, 48, 49). Patients learned the ABC(DE)s of their emotional disturbance, which consisted of identifying their activating events (A), their beliefs (B) about the activating events, and the resulting emotional and behavioral consequences (C). They learned to identify their irrational beliefs and to recognize the connection between the irrational beliefs and the emotional and behavioral disturbance, thus learning that the irrational beliefs could be replaced with rational beliefs to experience functional emotions and adaptive behaviors. Ellis’ elegant solution (48) was used in approaching patients’ beliefs, as the therapists focused on changing patients’ surface (e.g., it is awful that she does not love me) and core irrational beliefs (e.g., it is awful to not be loved by loved ones) and not automatic thoughts in the form of descriptive (e.g., my wife has another partner) or inferential cognitions (e.g., she does not love me). Patients’ irrational cognitions were approached by using empirical, pragmatic, and/or logical cognitive restructuring and/or debating strategies (D) that were based mainly on Socratic questioning in the context of a sound therapeutic relationship. Humor, metaphors, and spirituality- or religion-informed techniques were also used for cognitive restructuring when appropriate. Effective life philosophies (E), reflecting rational thinking, were elaborated on to replace the irrational beliefs. At the end of each session, homework assignments were negotiated in connection with the therapeutic goals, such as monitoring one’s mood by using the ABC(DE) self-help form, debating one’s irrational beliefs, rehearsing rational beliefs, and conducting behavioral experiments in the form of shame attacks. Behavioral strategies were used to target irrational beliefs, strengthen conviction in the rational beliefs, and/or target unconscious information processing (e.g., exposure to counter maladaptive classical conditioning). Problem-solving strategies were used to deal with the activating events, after the dysfunctional feelings and/or behaviors were changed to functional feelings and/or behaviors or during some crisis situations. Coping skills were also used to deal directly with consequences, especially in the crisis situations and for various homework assignments.

An alpha level of 0.05 was used for all statistical tests. Participants were included if they had completed at least one therapy session; participants who had participated in fewer than three sessions were considered to have dropped out of the study (38). Analyses concerning the outcome were conducted by using the intent-to-treat principle: the pairwise-comparison analysis included all patients, regardless of their protocol adherence or subsequent withdrawal from the study or assessments. The last available score on each outcome measure served as the termination score for those who dropped out of the study. We included analyses for early responders after three sessions of therapy (i.e., the early-response phase) and at the end of the 20 sessions (i.e., therapy termination). This decision was based on the literature (35) documenting the dose-effect relationship, which has shown that a positive early response to treatment can occur within the first three sessions, and research (40) supporting a good-enough level model of change. Effect size analyses for outcomes were based on Cohen’s d coefficients (50). We used Morris and DeShon’s (51) equation 8 to correct for dependence between means for within-subject data.

Treatment responses were also analyzed categorically, by using cutoff scores indicating symptoms of clinical significance and reliable change norms of the outcome (27). Reliable change is indicated when a patient’s score changes by 14 points or more on the OQ-45 total score. The following categories and criteria were used: “recovered” (patients with clinical scores at intake that improved and moved into the nonclinical range at termination, as indicated by the reliable change index), “improved clinical” (patients with clinical scores at intake that improved [reached a reliable change level at termination] but remained in the clinical range), “improved nonclinical” (patients whose scores were initially in the nonclinical range and registered reliable improvement at termination, with the score remaining in the nonclinical range), “deteriorated clinical” (patients with scores in the clinical range at intake that deteriorated with a reliable change level at termination), “deteriorated nonclinical” (patients whose scores were in the nonclinical range at intake but deteriorated with a reliable change level at termination), and “no reliable change” (patients whose scores did not register a reliable change level at termination). We used the same categories to analyze the recovery rates and clinically significant changes after the early-response phase (after three sessions) and at the termination phase (20th session).

Table 1 presents the means and standard deviations for the participants’ total scores on the OQ-45 and its subscales. At intake, 75% (N=262) of the participants registered a total OQ-45 score above the clinical cutoff score, with 70% (N=244) of the patients scoring in the clinical range for symptom distress, 69% (N=241) for social role functioning, and 74% (N=258) for interpersonal relationships.

Treatment dropout (i.e., completing fewer than three therapy sessions) was 18% (N=63). Of the participants, 73% (N=262) completed more than three sessions (24% [N=84] completed only three sessions), and 25% (N=87) completed 20 sessions. Patients attended a mean of 4.35±6.21 sessions, and 23% (N=80) attended more than 20 sessions (up to 112 sessions, with 9% [N=31] attending more than 40 sessions).

Descriptive statistics and comparisons between the total OQ-45 score at intake and the scores reported for each session are presented in Figure 2 and Table 2. We found significantly lower symptom severity for patients at the early-response phase (after three sessions), at termination (20 sessions), and at the last session, compared with the intake assessments for both the OQ-45 total scores and the OQ-45 subscales (all p<0.001) (Table 1).

Thirty-seven percent (N=129) of the patients had nonclinical total OQ-45 scores after the early-response phase, 46% (N=161) had nonclinical scores at termination, and 47% (N=164) had nonclinical scores after their last session of therapy. Of the patients, 34% (N=119) registered reliable change improvements after the early-response phase, 49% registered reliable change at termination (N=171), and 50% at their last session (N=174). Table 3 presents the recovery status of the participants after the early-response phase and at termination. Significant changes were found in clinical status of the patients’ total OQ-45 scores at intake compared with after the early-response phase (χ²=69.18, df=1,349, p<0.001, ε²=0.44) and at the termination of therapy (χ²=94.35, df=1,349, p<0.001, ε²=0.52).

We conducted exploratory correlation analyses between symptom severity and the number of therapy sessions attended. Significant correlation was found between the number of sessions and symptom severity at the end of treatment (r=–0.582, df=349, p<0.001; values were similar for clinical [r=–0.588, df=262, p<0.001] and nonclinical [r=–0.577, df=87, p<0.001] participants).