THERAPIST VARIABLES THAT PREDICT SYMPTOMCHANGE IN PSYCHOTHERAPY WITH CHRONICALLY

DEPRESSED OUTPATIENTS

CARINA VOCISANOSouthern Connecticut State University

DANIEL N. KLEINState University of New York at Stony Brook

BRUCE ARNOWStanford University

CARIDAD RIVERASouthern Connecticut State University

JANICE A. BLALOCKUniversity of Texas

BARBARA ROTHBAUMEmory University School of Medicine

DINA VIVIANState University of New York at Stony Brook

JOHN C. MARKOWITZ ANDJAMES H. KOCSIS

Cornell University Medical Center

RACHEL MANBERStanford University

LOUIS CASTONGUAYPennsylvania State University

A. JOHN RUSHUniversity of Texas Southwestern

Medical Center

FRANCES E. BORIANBristol-Myers Squibb

JAMES P. MCCULLOUGH ANDSUSAN G. KORNSTEIN

Virginia Commonwealth University

LAWRENCE P. RISOGeorgia State University

MICHAEL E. THASEUniversity of Pittsburgh Medical Center

Carina Vocisano and Caridad Rivera, Psychology Depart-ment, Southern Connecticut State University; Daniel N.Klein, Departments of Psychology and Psychiatry and Behav-ioral Sciences, State University of New York at Stony Brook;Bruce Arnow and Rachel Manber, Department of Psychiatryand Behavioral Sciences, Stanford University; Janice A.Blalock, M. D. Anderson Cancer Center, University of Texas;Barbara Rothbaum, Department of Psychiatry, Emory Uni-versity School of Medicine; Dina Vivian, Department of Psy-chology, State University of New York at Stony Brook; JohnC. Markowitz and James H. Kocsis, Department of Psychia-try, Cornell University Medical Center; Louis Castonguay,

Department of Psychology, Pennsylvania State University; A.John Rush, Department of Psychiatry, University of TexasSouthwestern Medical Center; Frances E. Borian, Bristol-MeyersSquibb, Princeton, New Jersey; James P. McCullough and SusanG. Kornstein, Department of Psychology, Virginia Common-wealth University; Lawrence P. Riso, Department of Psychol-ogy, Georgia State University; Michael E. Thase, Adult Aca-demic Psychiatry, University of Pittsburgh Medical Center.

Correspondence concerning this article should be ad-dressed to Carina Vocisano, Southern Connecticut State Uni-versity, 501 Crescent Street, New Haven, CT 06515-1355.E-mail: vocisanoc2@southernct.edu

Psychotherapy: Theory, Research, Practice, Training Copyright 2004 by the Educational Publishing Foundation2004, Vol. 41, No. 3, 255–265 0033-3204/04/$12.00 DOI 10.1037/0033-3204.41.3.255

255

The present study examines thecontribution of psychotherapistvariables to change in depressivesymptoms in a large clinical trialcomparing the efficacy of thecognitive–behavioral analysis system ofpsychotherapy, the antidepressantnefazodone, and the combination ofboth in the treatment of chronicdepression. Greater change on theHamilton Rating Scale for Depression(HRSD) was associated with greateremphasis on the therapeuticrelationship, lower overallpsychotherapy caseload, therapistpsychodynamic orientation, andsupervisory status. There was norelationship between HRSD change andtherapist sex, age, or years ofexperience.

Over the past decade, there has been consider-able concern among psychotherapy researchersover the neglect of therapist variables in psycho-therapy research (Beutler, 1997). The largest andmost influential psychotherapy outcome studieshave focused primarily on technique efficacy(Bergin, 1997), despite the fact that therapy out-come is more closely related to therapist charac-teristics than to type of treatment (Lambert &Okiishi, 1997; Luborsky et al., 1986). The large-scale and influential studies that have giventherapist characteristics some attention confirmthat therapists vary substantially in their successeven when they are working with patients whoare all being treated for the same disorder withthe same form of carefully monitored, manual-ized treatment (Blatt, Sanislow, Zuroff, & Pilko-nis, 1996; Huppert et al., 2001).

Huppert and colleagues (2001) conducted astudy of therapist characteristics and treatmentoutcome based on data from the Multicenter Col-laborative Study for the Treatment of Panic Dis-order. In the conclusion of their study, Huppertand colleagues (2001) pointed out that an idealstudy of therapist effects on therapy outcomewould include all patients who are receiving psy-chotherapy alone (as opposed to some patientswho are receiving psychotherapy and medica-

tions), a large number of clinicians, and a suffi-cient number of patients per therapist. The pres-ent study evidences all of these characteristics.We examine the relationship of a number of im-portant psychotherapist variables to therapy out-come in a large sample of outpatients withchronic depression who completed the psycho-therapy alone condition of a multisite clinicaltrial.

Our research looks at the traditional therapistvariables that have been included in most studiesconcerned with the impact of therapist variableson patient outcome, such as sex, age, and profes-sional experience. The present study also contrib-utes to the literature on these variables by pro-viding information derived from a large patientsample, a large therapist sample, and a rigorousresearch design. The present study also evaluatesthe impact of several important therapist vari-ables that have never been researched, such asoverall patient caseload across all settings, globalintervention strategy, and supervisory status.

Bowman, Scogin, Floyd, and McKendree-Smith (2001) used meta-analysis to summarize58 studies on the relationship of therapist sex totreatment outcome and found a statistically sig-nificant but small advantage for female thera-pists. They found no differences in outcome as-sociated with patient sex.

Therapy researchers have given little attentionto therapist age because it is confounded bytherapist experience and cohort effects that re-flect changes in training and standards of prac-tice. Beck (1988), however, suggested that thera-pists who are more than 10 years younger thantheir clients obtained poorer outcomes than oldertherapists or those of similar age to their clients.

Ethnic similarity between therapists and clientshas been shown to predict therapy outcome in afew studies (Hosch et al., 1995; Ricker, Nystul, &Waldo, 1999; Sue, Fujino, Hu, Takeuchi, &Zane, 1991). The large-scale clinical trials thatexamined the relationship between therapist char-acteristics and treatment outcome have not in-cluded information on therapist ethnicity (Blatt etal., 1996; Huppert et al., 2001), although the lat-ter study did report that their study patients werepredominantly Caucasian. In the present study,the vast majority of both therapists and clientsdescribed themselves as Caucasian, so we had toofew ethnic minority group members to conduct ameaningful analysis of the relation betweentherapist ethnicity and patient outcome. Future

Vocisano et al.

256

clinical trials should seek to ensure an adequaterepresentation of therapists and clients who aremembers of minority groups.

A similar situation occurred with respect toprofessional discipline; there was not enoughvariability in our clinician sample to discernwhether discipline related to therapy outcome (al-most all of the therapists were psychologists).The results of the only meta-analysis on this vari-able (Smith, Glass, & Miller, 1980) found a slightpatient outcome advantage when therapy wasconducted by a psychologist rather than a psy-chiatrist. A Consumer Reports (1995) study onclient-rated self-improvement showed no differ-ences between these two groups and showedgreater improvement in clients seen by socialworkers.

The relationship between level of professionalexperience and treatment outcome has been re-viewed in several meta-analyses with mixed con-clusions (Christensen & Jacobson, 1994; Stein &Lambert, 1995). Some of the difficulties in un-derstanding the relationship of professional expe-rience to therapy outcome are due to the variedand problematic ways that this variable has beenoperationally defined (Beutler, 1997). In keepingwith Beutler’s recommendation regarding themeasurement of therapist experience, we askedabout the number of years during which thera-pists conducted individual psychotherapy ratherthan the amount of time that had passed sincethey assumed the professional role.

There is a plethora of research that attempts todiscern the relation between specific interven-tions and patient change through observation andratings of therapist in-session behavior. We wereinterested in knowing whether therapists’ ownperceptions of the degree of emphasis they placeon specific interventions related to treatment out-come. Thus, we asked therapists to report on therelative emphasis they placed on a variety of in-terventions (global intervention strategy) to seewhether intervention emphasis was related totherapy outcome.

The influence of clinicians’ overall caseloadon patient care is frequently discussed in the lit-erature on professional burnout but has neverbeen studied in direct relation to patient outcome.Burnout is a syndrome of emotional exhaustionand cynicism that frequently occurs as a result of“constant or repeated emotional pressure associ-ated with an intense involvement with peopleover long periods of time” (Pines, Aronson, &

Kafry, 1981, p.15). Burnout involves a loss ofconcern and positive feeling for one’s clients anda decline in the quality of service that the clientsreceive (Edelwich & Brodsky, 1980). Given thetheoretical literature on professional burnout, wewere very curious about whether clinicians withhigh caseloads across all of their work settingshad clients within the study who evidenced lesschange.

We have also included two variables whoseinfluence may be unique within the context ofcontrolled clinical trials: therapists’ self-identified primary theoretical orientation and su-pervisory status. It seems reasonable to assumethat therapists’ primary theoretical orientation ex-erts some influence on how manualized therapyis interpreted and delivered in controlled clinicaltrials. For example, a therapist whose primarytheoretical orientation is psychodynamic may ap-ply manualized cognitive–behavioral therapy(CBT) differently than one whose primary theo-retical orientation is cognitive–behavioral. Thera-pists’ self-identified primary orientation hasrarely been considered when the outcomes of dif-ferent forms of manualized psychotherapy arecompared with one another and/or with medica-tions. Huppert and his colleagues (2001) did ex-amine the relationship of theoretical orientationto outcome and found none, but their study in-cluded only 14 therapists who could identifythemselves as cognitive–behavioral therapists or“other.” In controlled clinical trials, clinicianswho supervise often serve as study therapists aswell. Despite this, the possible relationship be-tween therapists’ supervisory status and treat-ment outcome has not been considered in pasttrials.

Method

Study Design

The present study is based on data from alarger study comparing the efficacy of the cogni-tive–behavioral analysis system of psychotherapy(CBASP; McCullough, 2000), the antidepressantnefazodone, and combination treatment in the 12-week acute phase in a multisite clinical trial fortreatment of chronic depression (Keller et al.,2000). The methods employed in this multisiteclinical trial are described in detail by Keller andhis colleagues (2000); hence, only a brief sum-mary is provided here. A total of 681 patients at

Therapist Variables That Predict Change

257

12 academic centers were randomized to 12weeks of treatment with CBASP alone, nefaz-odone alone, or the combination of both treat-ments. In that study, 52% of patients who com-pleted psychotherapy alone, 55% of patients whocompleted nefazodone alone, and 85% of patientswho completed combination treatment respondedto treatment (Keller et al., 2000). Thus, the twomonotherapies had equivalent efficacy, and com-bination treatment produced significantly betteroutcomes than did either treatment alone. Thepresent report is based on a sample of 173 pa-tients who completed psychotherapy alone with12 weeks of acute treatment.

All patients, therapists, and supervisors pro-vided written informed consent. The psycho-therapists attended a 2-day training workshop andmet the criteria for mastery of treatment proce-dures entailed in CBASP, as assessed by evalu-ation of their performance during two videotapedpilot cases. During the study, all psychotherapysessions were videotaped, and supervisors re-viewed a sample of the videotapes weekly to dis-cuss the therapeutic exchange and to assess thepsychotherapists’ adherence to CBASP. Supervi-sion was conducted in group and/or individualformats. Most therapists received at least oneweekly group supervision (69.2%), but a signifi-cant minority also received supervision individu-ally (23.1%). Only 2 therapists received indi-vidual supervision without group supervision(7.7%).

CBASP

CBASP was designed by James McCullough(2000) to motivate and help chronically de-pressed patients to develop more effective socialproblem-solving and relationship skills. Situ-ational analysis is the primary tool for helpingpatients to change their patterns of coping, im-prove their interpersonal skills, understand theconsequences of their behavior, interact more ef-fectively with others, and expand their social sup-port network. During situational analysis, the cli-ent focuses on one recent interpersonal situationthat is distressing and deconstructs the compo-nents of the situation with the therapist. Situ-ational analysis has three phases: elicitation, re-mediation, and generalization. In the elicitationphase, the patient describes (a) the interpersonalevent, (b) his or her behavior, (c) his or her in-terpretations of what occurred, (d) the outcome of

the event (actual outcome), (e) what he or shewould have liked the outcome to be (desired out-come), and (f) why he or she did or did notachieve the desired outcome. In the remediationphase, the client’s interpretations and behaviorsduring the event are analyzed and hypotheticallyrevised in terms of what would have been morelikely to bring about the patient’s desired out-come. In this manner, the patient is assisted inexamining the consequences of his or herthoughts and behavior and in constructing alter-native ways of thinking and behaving that wouldhave led to more effective social problem solv-ing. Generalization involves a review of what hasbeen learned in the preceding analysis and anexploration of the ways the patient’s new under-standing and skills may be applied to similar situ-ations in the future. CBASP also provides aframework for therapists to use the patient–therapist relationship as a vehicle for improvingproblem-solving skills in the interpersonal do-main. During the second session of CBASP, ahistory of the client’s relationships with signifi-cant others is elicited. This information is used toidentify and address patterns of negative interper-sonal expectations that may be enacted in thetherapeutic relationship and in the patient’s life.

The psychotherapists followed a manual (Mc-Cullough, 1995) and a study protocol specifyingtwice-weekly sessions during Weeks 1 through 4and weekly sessions during Weeks 5 through 12.Twice-weekly sessions could be extended untilWeek 8 if a patient was not adequately perform-ing the social problem-solving procedure accord-ing to the criteria. Psychotherapy was extended toWeek 14 if the patient’s treatment response wasequivocal. All psychotherapy sessions were vid-eotaped, and a sample was reviewed weekly bysupervisors.

Participants

Participants in the present study included threedistinct groups: (a) patients who underwent treat-ment as part of the chronic depression treatmentstudy, (b) psychotherapists who conducted thepsychotherapy with these patients and, (c) psy-chotherapy supervisors, all of whom also servedas psychotherapists in the study.

Patient participants. Patients were recruitedat 12 academic sites between June 1996 and De-cember 1997. They fulfilled criteria for one ofthree forms of chronic depression: (a) current ma-

Vocisano et al.

258

jor depressive episode of at least 2 years’ dura-tion, (b) current major depressive episode super-imposed on a preexisting dysthymic disorder(double depression), (c) recurrent major depres-sive disorder with incomplete remission betweenepisodes, a current major depressive episode, anda continuous illness of at least 2 years. Diagnoseswere derived using the Structured Clinical Inter-view for Axis I DSM–IV (American PsychiatricAssociation, 1994) Disorders (SCID-I/P; First,Spitzer, Gibbon, & Williams, 1995).

To be eligible for the study, the patients had tobe between the ages of 18 and 75 years and tohave a score of at least 20 on the 24-item Ham-ilton Rating Scale for Depression (HRSD; Ham-ilton, 1967) at screening and after a 2-week drug-free baseline period. The exclusion criteria in-cluded presence of neurological disorders, aserious medical condition, or a history of psycho-sis, mania/hypomania, or substance-related disor-ders. Patients unresponsive to several previousadequate trials of antidepressant medication andempirically supported forms of psychotherapy fordepression were also excluded (for a detailed de-scription of study exclusion criteria, see Keller etal., 2000).

To avoid the confounding effects of antide-pressant medications on patient outcomes, in thepresent analyses we included only patients re-ceiving psychotherapy without medications. Datafor the study were available for all of the patientswho received psychotherapy alone and com-pleted the 12-week acute phase (N � 173). Su-pervisors were included in the analyses as thera-pists, except where noted.

The majority of the 173 patients were female(63%) and Caucasian (90.8%). Their average agewas 43.9 years (SD � 10.6). Almost half of thepatients were married (36.4%) or cohabiting(6.9%), approximately one third (31.8%) weresingle, and about a quarter were divorced(19.1%), separated (4.6%), or widowed (1.2%).

The distribution of chronic depressive disor-ders in this sample (N � 173) was chronic majordepression (35.3%), recurrent major depressionwith incomplete recovery (23.7%), double de-pression (23.1%), and chronic major depressionsuperimposed on dysthymia (17.9%). The aver-age baseline HRSD score was 26.5 (SD � 4.9),and the Global Assessment of Functioning Scale(Axis V in the DSM–IV) scores averaged 53.8(SD � 5.6). The average age of onset of majordepressive disorder was 27.4 years (SD � 13.1),

and the mean duration of the current major de-pressive episode was 8.5 years (SD � 10.7). Theaverage age of onset of dysthymic disorder was20.5 years (SD � 15.0), with a mean duration of21.9 years (SD � 15.0). Most patients had beenin psychotherapy (63.0%) or had been treatedwith antidepressants (58.4%) prior to participat-ing in the study.

Therapist and supervisor participants. A de-tailed description of therapist demographic andprofessional characteristics are presented in theResults section and in Table 1. All of the 40psychotherapists and 12 psychotherapy supervi-sors involved in the treatment study were invitedto participate in this study. Three therapists and 1supervisor conducted therapy with patients incombination treatment but did not conducttherapy with any patients in psychotherapy alone.We did not include these therapists and supervi-sors in the data analyses. Thus, we have data on37 therapists and 11 supervisors. Two supervisorsand 2 therapists had incomplete data and are notincluded in some analyses. Psychotherapy super-visors conducted psychotherapy with study pa-tients and were included as psychotherapists inthe analyses where appropriate. Study therapistswere required to have had at least 2 years expe-rience after earning an MD or PhD or at least 5years of experience after earning an MSW. Allpsychotherapists had a PhD in psychology, ex-cept for 3 clinicians who had an MSW and 1 whohad an MD with a specialty in psychiatry.

Study supervisors were required to have had atleast 7 years postdoctoral or postmedical degreeor 10 years post-MSW experience and a cogni-tive–behavioral theoretical orientation. All super-visors had a PhD in psychology, except for 1 whohad an MSW.

Measures

Diagnoses were derived using a modified ver-sion of the Structured Clinical Interview for theDSM–IV (SCID I/P; First, Spitzer, Gibbon, &Williams, 1995) for Axis I disorders and an ab-breviated version of the SCID–II (First, Gibbon,Spitzer, Williams, & Benjamin, 1997) for Axis IIdisorders during the screening evaluation. Aphysical examination, routine laboratory tests,and urine toxicology screen were also performedat this time.

Our primary outcome measure was the 24-itemHRSD, which was administered at screening,baseline, and Weeks 1, 2, 3, 4, 6, 8, 10, and 12 by

Therapist Variables That Predict Change

259

raters who were unaware of patients’ treatmentconditions. SCID and HRSD raters completed a2-day training workshop on these instruments. Inaddition, all HRSD raters were certified by ex-ternal experts who evaluated a videotape of therater’s HRSD assessment with a depressed patient.The HRSD interviewers were blind to group as-signment and were certified to have a high levelof interrater reliability and procedural integrity.

We also developed a supplementary self-reportquestionnaire completed by therapists and super-visors that included quantitative and qualitativequestions to assess professional training andpractice experiences, total clinical caseload, re-search caseload, primary theoretical orientations,and intervention emphasis. We sought to developa self-report questionnaire that would (a) includethe main therapist variables that have been exam-ined in the therapy outcome literature, (b) takeinto account criticisms of how these variableshave been measured in previous studies, and (c)include some new and important variables.

In keeping with Beutler’s (1997) recommenda-tion regarding the measurement of therapy expe-rience, we asked therapists about the number ofyears during which they conducted individualpsychotherapy rather than the amount of time thathad passed since they assumed the professionalrole. Previous therapy outcome studies had notexamined total caseload. We therefore askedtherapists about the average number of hours per

week that they devoted to conducting psycho-therapy across all work settings. We also askedthem to describe their primary theoretical orien-tation in recognition of the fact that most clini-cians use more than one therapy approach. Al-though past research has examined therapists’ ob-served use of specific intervention strategiesduring psychotherapy sessions, we were inter-ested in therapists’ subjective perspective on theirrelative intervention emphasis. Thus, the inter-vention strategy items required therapists to re-port on the relative degree to which they empha-sized each of five listed interventions across allpatients in their research caseload: remediation ofinterpretations, remediation of behavior, modifi-cation of situational outcomes, examination ofthe relationship between patients’ social historiesand their current behavior, and discussions of thepatient–therapist relationship. Therapists ratedthe percentage of emphasis they placed on eachof these five interventions, such that the totaladded up to 100%.

Data Analysis

The main dependent variable was the mean ofthe change scores from baseline to end of theacute phase on the HRSD for all patients treatedby each therapist. On average, therapists workedwith 3 study patients at a time (M � 3.2, SD �2.3), but a couple of therapists had much highercaseloads (range � 0–12). Thus, the mean of the

TABLE 1. Therapist Sex, Age, Experience, Clinical Caseload in the CBASP Study, and Clinical Caseload AcrossAll Settings

VariableTherapists(n � 37)

Supervisors(n � 11)

All therapists(n � 48)

Correlations with changein HRSD scores

Female(%) 25 (67.6%) 5 (45.5%) 30 (62.5%)a

AgeM 44.0 49.4 45.1 r � .03SD 8.1 10.1 8.7

Years of experience as psychotherapistM 15.3 20.7 16.4 r � −.11SD 7.8 9.9 8.4

Caseload in the CBASP studyb

M 3.8 1.5 3.4 r � .05SD 2.3 0.5 2.4

Caseload across all settingsc

M 21.9 16.2 20.4 r � −.47*SD 12.4 6.9 11.8

Note. CBASP � cognitive behavioral analysis system of psychotherapy; HRSD � Hamilton Rating Scale for Depression.aThere were no differences in mean HRSD change scores between patients of female therapists (M � 10.19, SD � 6.0) versusmale therapists (M � 12.85, SD � 5.5). bNumber of hours per week conducting therapy in the CBASP study. cNumber ofhours per week conducting therapy across all settings.*p < .002.

Vocisano et al.

260

HRSD change scores for all of the patients seenby each therapist within the study was the maindependent measure. Although the use of rawchange scores has been criticized in the past,more recent work has demonstrated that thesecriticisms were often overstated and that the useof residual change scores is often equally prob-lematic (Rogosa, 1995; Willet, 1988).

We performed Pearson product–moment cor-relations to analyze the relationship between con-tinuous therapist variables and patient outcome.We used univariate and multivariate analyses ofvariance to discern possible group differences be-tween CBT supervisors, CBT-oriented therapists,and psychodynamic therapists, and we performedpost hoc Tukey honestly significant difference(HSD) analyses where appropriate. We con-ducted a simultaneous multiple regression analy-sis to examine the independent contributions ofsupervisory status/therapeutic orientation, psy-chotherapy caseload, and intervention strategy topredicting patient outcome.

Results

In the first section, we present data on the re-lationship of several key therapist variables topatient change, including therapist age, sex, num-ber of years of experience practicing individualpsychotherapy, clinical caseload in the researchstudy, and clinical caseload across all settings.We also present data on the relationship betweenself-identified primary theoretical orientation andsupervisory status to outcome. Finally, we pre-sent our findings on the relationship betweentherapist global self-report about the use of inter-vention strategies and patient outcome.

The Relationship of Therapist Variables toPatient Outcome

Data on therapist sex, age, experience, case-load, and primary therapeutic orientation are pre-

sented in Tables 1 and 2. There were no differ-ences in mean HRSD change scores between pa-tients of female (M � 10.19, SD � 6.0) versusmale therapists (M � 12.85, SD � 5.5). Wecould not meaningfully address how patient andtherapist ethnicity or therapist discipline and pa-tient change relate because the vast majority oftherapists were psychologists who describedthemselves as Caucasian. Therapist age, the num-ber of years that a therapist had devoted to con-ducting psychotherapy (including training), andthe number of hours per week therapists spentworking with study patients were not correlatedwith change in depression scores (r � −.03, p <.87; r � −.11, p < .51; r � .05, p < .73, respec-tively). The average number of hours per weekthat psychotherapists and supervisors devoted topsychotherapy across all settings, however, wasnegatively correlated with patient change (r �−.47, p < .002).

As all supervisors had a cognitive–behavioralorientation, primary theoretical orientation andsupervisory status were combined to createa three-level dummy variable: cognitive–behavioral supervisor, cognitive–behavioraltherapist, and psychodynamic therapist. Thera-pists with other theoretical orientations were ex-cluded from these analyses (n � 5, 13.5%). Aunivariate analysis of variance yielded significantdifferences between the cognitive–behavioral su-pervisors, the psychodynamic psychotherapists,and the cognitive–behaviorally oriented thera-pists on average HRSD change scores, F(2, 38)� 5.57, p < .001. A post hoc Tukey HSD analy-sis showed a significant reduction in depressionscores for patients of therapy supervisors andpsychodynamic therapists when compared withpatients of the cognitive–behavioral psychothera-pists. There were no differences in change scoresbetween patients whose therapists were supervi-sors versus patients whose therapists identified

TABLE 2. Change in Hamilton Rating Scale for Depression (HRSD) Scores According to Supervisory Status andSelf-Identified Therapeutic Orientation

Change in HRSD-24Cognitive behavioral therapists

(excluding supervisors) (n � 20)Psychodynamic

therapists (n � 10)Cognitive behavioral

supervisorsa (n � 11)

M 9.3 14.1b 15.0c

SD 4.7 7.0 4.5

aAll of the supervisors were cognitive–behaviorally trained (CBT) because only CBT clinicians were selected to become studysupervisors. bTukey honestly significant difference, psychodynamic versus CBT therapists, p < .05. cTukey honestly signifi-cant difference, supervisors versus CBT therapists, p < .01.

Therapist Variables That Predict Change

261

their primary orientation as psychodynamic(Table 2).

A multivariate analysis of variance yielded sig-nificant differences among the three therapistgroups (CBT therapists, psychodynamic thera-pists, and supervisors) on global interventionstrategy, Wilks’s �(10, 62) � .493, p � .001.Follow-up univariate analyses of variance re-vealed significant differences among the threetherapist groups in emphasis on remediation ofinterpretations, F(2, 38) � 3.29, p < .05, anddiscussions of the patient–therapist relationship,F(2, 38) � 8.33, p < .001. Tukey HSD post hoctests showed that the supervisors placed signifi-cantly less emphasis on cognitive interventionsand greater emphasis on discussing the patient–therapist relationship as compared with CBTtherapists and that psychodynamic therapists alsoplaced greater emphasis on the therapist–patientrelationship than did CBT therapists. In addition,psychodynamic therapists placed significantlygreater emphasis on cognitive interventions thandid supervisors. (See Table 3.)

We also examined the relationship between in-tervention emphasis and symptom change. Agreater emphasis by therapists (including thosewho supervised) on the relationship betweentherapist and client was correlated with greaterchange in psychotherapy (Table 3).

We conducted a simultaneous multiple regres-sion analysis to examine the independent contri-butions of supervisory status/therapeutic orienta-tion (CBT supervisors, CBT therapists, psycho-dynamic therapists), overall psychotherapycaseload, and intervention strategy (emphasis onpatient–therapist relationship) to predicting pa-tient outcome. The multiple correlation betweenpatient outcome and the three predictor variableswas .61, F(3, 35) � 6.87, p < .001. The onlyunique predictor of change was emphasis on thepatient–therapist relationship (B � .37, R2 �.25, p < .02). Supervisor status/theoretical orien-tation and overall psychotherapy caseload did notmake independent contributions to outcome.None of the three predictor variables (supervisorystatus/therapeutic orientation, overall psycho-

TABLE 3. Therapist Global Intervention Strategy

Self-report of global intervention strategyCorrelations with

change in HRSD-24

CB therapists(no supervisors)

(n � 20)

Psychodynamictherapists(n � 10)

CB supervisors(n � 9)

Emphasis on cognitive interventionsR −.25p .11M% 30.2a 29.6 21.1SD 10.7 8.0 4.9

Emphasis on behavioral interventionsR −.30p .06M% 23.0 16.4 19.4SD 9.6 10.3 8.1

Emphasis on desired outcomesR .09p .55M% 21.2 19.4 21.1SD 5.9 4.6 4.2

Emphasis on patient–therapist relationshipR .48p .001M% 11.1b 22.0c 18.1SD 6.8 9.4 5.8

Relative emphasis on learning historyR .03p .85M% 15.0 12.8 16.9SD 5.6 5.7 6.4

Note. HRSD � Hamilton Rating Scale for Depression; CB � cognitive–behavioral.aTukey honestly significant difference (HSD), cognitive–behaviorally trained (CBT) therapists versus supervisors, p <.05. bTukey HSD, CBT therapists versus supervisors, p < .05. cTukey HSD, psychodynamic therapists versus CBT therapists,p < .001.

Vocisano et al.

262

therapy caseload across all settings, and emphasison the therapeutic relationship) were correlatedwith baseline HRSD scores (r � .06, p < .74;r � .00, p < .98; r � .06, p < .72, respectively).

We also conducted a series of analyses to ad-dress the possibility that the variability in thenumber of cases that each therapist treated couldinfluence the findings. We performed Spearmanrank order correlations to ascertain whether thekey independent variables (supervisory status/therapeutic orientation, overall caseload acrosssettings, and emphasis on the patient–therapistrelationship) were correlated with the total num-ber of study patients treated by each therapist andwith the average within-study caseload. None ofthe key independent variables were correlatedwith the total number of patients treated by eachtherapist or the average within-study caseload.

Discussion

The single best predictor of psychotherapy out-come in this study was the overall degree of em-phasis therapists placed on discussing the pa-tient–therapist relationship. The overall emphasistherapists report placing on remediation of cog-nitions, behaviors, and desired outcomes and onexamination of patient social learning history wasnot associated with symptom reduction. Symp-tom change was negatively correlated with thetotal number of hours therapists spent conductingpsychotherapy across all settings, but it was notrelated to therapist sex, age, experience, or case-load within the study.

Supervisory status and self-identified primaryorientation were also related to symptom change;patients of cognitive–behavioral psychotherapysupervisors and therapists who identified theirprimary orientation as psychodynamic evidencedsignificantly greater change on the HRSD ascompared with patients of cognitive–behavioralpsychotherapists.

The predictive value of the degree of emphasistherapists placed on the therapeutic relationshipis congruent with past research (Safran & Wall-ner, 1991) highlighting discussions of the thera-pist–client relationship as a key component ofchange in CBT. Our findings are also consistentwith past research showing that interventions tar-geting interpersonal domains in cognitive therapyfor depression are associated with improvement(Hayes, Castonguay, & Goldfried, 1996; Jones &Pulos, 1993).

The results of the present study suggest that thebest psychotherapy outcomes, at least for chroni-cally depressed patients, are facilitated by thera-pists who use a sophisticated blend of cognitive–behavioral and interpersonal/psychodynamic ap-proaches. A number of findings support thisperspective. First, the CBASP method, althoughit emphasizes cognitive–behavioral interventions,is composed of a blend of therapeutic approachesand has previously been demonstrated to be effi-cacious in treating chronic depression (Keller etal., 2000). Second, supervisors whose primaryorientation was cognitive–behavioral yielded out-comes that were equivalent with those of psycho-dynamic therapists, and both groups reported us-ing the most blended intervention strategies. Fi-nally, cognitive–behavioral therapists who didnot supervise reported using the least integratedform of treatment and yielded outcomes that weresignificantly poorer than those achieved by psy-chodynamic therapists and cognitive–behavioralsupervisors. Our findings agree with those ofGoldfried, Raue, and Castonguay (1998). In theirstudy, the portions of sessions that master thera-pists judged as clinically significant reflected ablending of both psychodynamic and cognitive–behavioral orientations, regardless of the primaryorientation of the therapist.

In interpreting these results, it is important toconsider that although the supervisors weresomewhat more experienced than psychothera-pists who did not supervise, the number of yearsthat a clinician had practiced individual psycho-therapy was not related to patient outcome. Thesuperior performance of the clinical supervisors(relative to the CBT supervisees) cannot, there-fore, be attributed to greater experience in con-ducting individual psychotherapy.

One possible explanation for the lack of asso-ciation between therapist experience and patientoutcome concerns the relevance of past experi-ence in this study. Experience may not have beenrelated to patient outcome because CBASP wasnew to almost all therapists and supervisors. Therelationship between therapist experience andtherapy outcome may have also been limited bythe fact that all therapists were quite experienced,having practiced psychotherapy for an average ofmore than 16 years.

We found that the number of hours therapistsspent treating chronically depressed patients inthe research study was not related to patient out-come—a finding congruent with the results of the

Therapist Variables That Predict Change

263

Treatment of Depression Collaborative ResearchProgram (Elkin, 1999). However, the inverse re-lationship between the total number of hours cli-nicians devoted to psychotherapy across all set-tings and patient outcome suggests that too manyconcentrated hours of psychotherapy may taxtherapists and negatively impact patients as well.Although there has been ample concern in thefield about therapist burnout, we were surprisedto find no previous research on the relationshipbetween overall psychotherapy caseload andpatient outcome. More research on this issue isneeded, as it has critical public health implications.

Given the small number of therapists, espe-cially in the supervisor and psychodynamicgroups, there was a moderate risk of Type II er-ror, and it is likely that results could be detectedonly if the effect sizes were relatively large. Wedid not, therefore, apply a correction for the num-bers of analyses that we ran, and this increasedthe risk of Type I error. In addition, some datawere missing from some therapists and supervi-sors on some variables. For the above reasons,the results of this study should be considered pre-liminary. Another limitation of this study con-cerns the fact that the therapist and supervisordata were collected on a one-time basis, when theacute phase of the project was nearly complete.Thus, the therapists and supervisors provided ret-rospective self-reports, and the results of thestudy need to be interpreted in this context. Al-though self-report offers a distinct and meaning-ful perspective on the therapeutic relationship, itdoes not necessarily correlate with in-session be-havior (Carroll, Nich, & Rounsaville, 1998). Inaddition, some measures were developed solelyfor this study and have unknown psychometricproperties. Finally, the results of this study maynot generalize to nonacademic settings or to otherpatient populations.

The present study has several strengths. Over-all, a larger number of patients, therapists, andsupervisors participated in this investigation, ascompared with past studies on the relationship oftherapist characteristics to therapy outcome. Con-sequently, few psychotherapy studies have hadcomparable statistical power. In addition, we as-sessed the multiple perspectives of independentevaluators, patients, therapists, and supervisors.This decreased the influence of demand charac-teristics for evidence of improvement and thepossible impact of biases about treatments orclinicians.

If replicated, the results of this study have im-portant implications for the organization and de-livery of psychotherapy services. An integratedapproach with ample discussion of the therapeu-tic relationship and a limit on the total number ofhours psychotherapists spend conducting therapyare important ingredients of treatment that sub-stantially reduce symptoms of depression.

References

American Psychiatric Association. (1994). Diagnostic andstatistical manual of mental disorders (4th ed.). Washing-ton, DC: Author.

BECK, D. F. (1988). Counselor characteristics: How they af-fect outcomes. Milwaukee, WI: Family Service of America.

BERGIN, A. (1997). Neglect of the therapist and the humandimensions of change: A commentary. Clinical Psychol-ogy: Science and Practice, 4, 83–89.

BEUTLER, L. E. (1997). The psychotherapist as a neglectedvariable in psychotherapy: An illustration by reference tothe role of therapist experience and training. Clinical Psy-chology: Science and Practice, 4, 44–52.

BLATT, S. J., SANISLOW, C. A., ZUROFF, D., & PILKONIS, P.(1996). Characteristics of the effective therapist: Furtheranalysis of the data from the NIMH TDCRP. Journal ofConsulting and Clinical Psychology, 64, 1276–1284.

BOWMAN, D. G., SCOGIN, F., FLOYD, M., & MCKENDREE-SMITH, N. (2001). Effect of therapist sex on outcome ofpsychotherapy: A meta-analysis. Psychotherapy, 38,142–148.

CARROLL, K. M., NICH, C., & ROUNSAVILLE, B. J. (1998).Utility of therapist checklists to monitor delivery of copingskills treatment for cocaine abusers. Psychotherapy Re-search, 8, 307–320.

CHRISTENSEN, A., & JACOBSON, N. D. (1994). Who (or what)can do psychotherapy: The status and challenge of nonpro-fessional therapies. Psychological Science, 5, 8–14.

Consumer Reports. (1995, November). Mental health: Doestherapy help? pp. 734–739.

EDELWICH, J., & BRODSKY, A. (1980). Burn-out: Stages ofdisillusionment in the helping professions. New York: Hu-man Sciences Press.

ELKIN, I. (1999). A major dilemma in psychotherapy outcomeresearch: Disentangling therapists from therapies. ClinicalPsychology: Science and Practice, 6, 10–32.

FIRST, M. B., GIBBON, M., SPITZER, R. L., WILLIAMS, J. B. W.,& BENJAMIN, L. S. (1997). Structured Clinical Interview forDSM–IV Axis II Disorders (SCID-II). Washington, DC:American Psychiatric Press.

FIRST, M. B., SPITZER, R. L., GIBBON, M., & WILLIAMS,J. B. W. (1995). Structured Clinical Interview for DSM-IVAxis I Disorders—Patient Edition (SCID-I/P, Version 2.0).New York: Biometrics Research Department, New YorkState Psychiatric Institute.

GOLDFRIED, M. R., RAUE, P. J., & CASTONGUAY, L. G. (1998).The therapeutic focus in significant sessions of mastertherapists: A comparison of cognitive–behavioral and psy-chodynamic–interpersonal interventions. Journal of Con-sulting and Clinical Psychology, 66, 803–810.

HAMILTON, M. (1967). Development of a rating scale for pri-

Vocisano et al.

264

mary depressive illness. British Journal of Social andClinical Psychology, 6, 278–296.

HAYES, A. M., CASTONGUAY, L. G., & GOLDFRIED, M. R.(1996). Effectiveness of targeting the vulnerability factorsof depression in cognitive therapy. Journal of Consultingand Clinical Psychology, 64, 623–627.

HOSCH, H. M., BARRIENTOS, G. A., FIERRO, C., RAMIREZ, J. I.,PELAEZ, M. P., CEDILLOS, M. A., ET AL. (1995). Predictingadherence to medications by Hispanics with schizophrenia.Hispanic Journal of Behavioral Sciences, 17, 320–333.

HUPPERT, J., BUFKA, L. F., BARLOW, D. H., GORMAN, J. M.,SHEAR, M. K., & WOODS, S. W. (2001). Therapists, thera-pist variables, and cognitive–behavioral therapy outcome ina multicenter trial for panic disorder. Journal of Consultingand Clinical Psychology, 69, 747–755.

JONES, E. E., & PULOS, S. M. (1993). Comparing the processof psychodynamic and cognitive–behavioral therapies.Journal of Consulting and Clinical Psychology, 61,306–316.

KELLER, M. B., MCCULLOUGH, J. P., KLEIN, D. N., ARNOW, B.,DUNNER, D. L., GELENBERG, A. J., ET AL. (2000). A com-parison of Nefazodone, the cognitive behavioral-analysissystem of psychotherapy, and their combination for thetreatment of chronic depression. New England Journal ofMedicine, 342, 1462–1470.

LAMBERT, M., & OKIISHI, J. C. (1997). The effects of theindividual psychotherapist and implications for future re-search. Clinical Psychology: Science and Practice, 4,66–75.

LUBORSKY, L., CRITS-CRISTOPH, P., MCLELLAN, T., WOODY,G., PIPER, W., LIVERMAN, B., ET AL. (1986). Do therapistsvary much in their success? Findings from four outcomestudies. American Journal of Orthopsychiatry, 56,501–512.

MCCULLOUGH, J. P. (1995). Therapist manual for cognitivebehavioral analysis system of psychotherapy. Richmond:Virginia Commonwealth University.

MCCULLOUGH, J. P. (2000). Treatment for chronic depres-sion: Cognitive-behavioral analysis system of psycho-therapy. New York: Guilford Press.

PINES, A. M., ARONSON, E., & KAFRY, D. (1981). Burnout:From tedium to personal growth. New York: Free Press.

RICKER, M., NYSTUL, M., & WALDO, M. (1999). Counselors’and clients’ ethnic similarity and therapeutic alliance intime-limited outcomes of counseling. Psychological Re-ports, 84, 674–676.

ROGOSA, D. (1995). Myths and methods: “Myths about lon-gitudinal research” plus supplemental questions. In J. M.Gottman (Ed.), The analysis of change (pp. 2–65).Mahwah, NJ: Erlbaum.

SAFRAN, J. D., & WALLNER, L. K. (1991). The relative pre-dictive validity of two therapeutic alliance measures in cog-nitive therapy. Psychological Assessment, 31, 188–195.

SMITH, M. L., GLASS, G. V., & MILLER, T. I. (1980). Thebenefits of psychotherapy. Baltimore: Johns Hopkins Uni-versity Press.

STEIN, D. M., & LAMBERT, M. J. (1995). Graduate training inpsychotherapy: Are therapy outcomes enhanced? Journalof Consulting and Clinical Psychology, 63, 182–196.

SUE, S., FUJINO, D. C., HU, L.-T., TAKEUCHI, D. T., & ZANE,N. W. S. (1991). Community mental health services forethnic minority groups: A test of the cultural responsive-ness hypothesis. Journal of Consulting and Clinical Psy-chology, 59, 533–540.

WILLET, J. B. (1988). Questions and answers in the measure-ment of change. Review of Research in Education, 15,345–422.

Therapist Variables That Predict Change

265