Computer programs for psychotherapy have been recommended as methods for making effective treatments readily available to more people (1–3). Potential benefits could include reduced health care costs, improved access to psychotherapy, enhanced delivery of psychoeducational information, and promotion of self-help.
Cognitive therapy for depression has been tested in many controlled studies (4, 5) and meets the standards of an empirically validated treatment (6). However, there is a paucity of appropriately trained therapists to provide standard cognitive therapy to the large numbers of patients with depression. The amount of clinician time required for treatment has typically ranged from 8 to 20 hours (4, 7).
One strategy for reducing clinician time has been the development of computer programs for psychotherapy. The first computer program for depression was developed in the 1980s by Selmi et al. (8). A randomized, controlled trial found that this program was as effective as standard cognitive therapy for mild to moderate depression. Because the software by Selmi et al. relies solely on text presentations and has not been updated, it is not used in clinical practice. Colby and Colby (9, 10) introduced a computer program in 1990 that uses written text to provide information on depression and a module that simulates dialogue between a therapist and a patient. A small controlled study of the program by Colby and Colby with depressed inpatients found that use of this software did not improve treatment outcomes (11).
The software that was tested in the current study was developed as an alternative to these earlier forms of computer-assisted therapy (12, 13). A multimedia format is used to engage patients, teach core methods of standard cognitive therapy, and reinforce learning. The computer program contains a variety of interactive self-help exercises designed to build skills for using cognitive and behavioral therapy. Video, audio, graphics, and checklists are used extensively throughout the program.
Our goal in this study was to demonstrate that computer-assisted standard cognitive therapy is effective in treating depression while reducing clinician time.
Medication-free outpatients with nonpsychotic major depression were randomly assigned to 8 weeks of computer-assisted cognitive therapy or standard cognitive therapy or to a wait list control group. Block randomization equated assignment of participants with a secondary diagnoses of dysthymia. All participants were treated at the Norton Psychiatric Center, which is affiliated with the University of Louisville.
Treatment with computer-assisted cognitive therapy included nine sessions with a therapist (first session=50 minutes, subsequent sessions=25 minutes) and eight computer sessions (20–30 minutes) that followed immediately after sessions 1–9. Standard cognitive therapy was delivered in nine sessions with a therapist (50 minutes). Selection of an 8-week format is consistent with the mean treatment duration reported by Dobson (4) for his meta-analysis of studies of standard cognitive therapy for depression (for nongeriatric studies of individual standard cognitive therapy, mean=9.9; for studies in which standard cognitive therapy was compared to a wait list, mean=6.6); it is also consistent with our goal to test the feasibility of a brief form of computer-assisted cognitive therapy as a strategy to improve the efficiency of treatment.
Assessments were completed by independent raters before treatment, after 4 and 8 weeks of therapy, and 3 and 6 months posttreatment. Study completers were those who attended at least six therapy sessions. Treatment, if any, in the follow-up period was not controlled or measured.
Participants (N=45) included women and men, ages 18–65, recruited by advertisements and by referral, who were screened for inclusion by a research nurse using the Structured Clinical Interview for DSM-IV Axis I (SCID) (14). Diagnoses were confirmed by interview with a clinical psychologist. Additional inclusion criteria were a 10th-grade education or General Equivalency Diploma, a ninth-grade reading ability on the Wide-Range Achievement Test (15), and a Beck Depression Inventory (16) score of at least 14.
Exclusion criteria were the following:
A diagnosis of schizophrenia or other psychotic disorder, bipolar disorder, depression secondary to substance abuse or medical condition, chronic major depression, obsessive-compulsive disorder, anorexia nervosa or bulimia nervosa, borderline personality disorder, dementia or other cognitive disorders, any substance use disorder other than nicotine, or anxiety disorder of greater severity than depression
Current suicidal ideation or plan (Beck Depression Inventory suicide item 2 or higher or clinician judgment) or a history of two suicide attempts or episodes of self-mutilation
Previous felony conviction, two arrests as a juvenile, or currently on probation
Past treatment with cognitive therapy
A medical condition that would interfere with participation in therapy
The study was approved by an institutional review board, and written informed consent was obtained from all participants.
Assessments were performed by independent, evaluators blind to the study design who were trained to administer the SCID-II and the 17-item Hamilton Depression Rating Scale (17). Interrater reliability on the Hamilton depression scale was 0.96 (intraclass correlation coefficient; see reference 18) before starting the study. Reliability was monitored by videotape at 3-month intervals throughout the investigation. Primary outcome measures were the Hamilton depression scale, the Beck Depression Inventory, the Automatic Thoughts Questionnaire (19), and the Dysfunctional Attitude Scale (20). Knowledge of cognitive therapy was measured with the Cognitive Therapy Awareness Scale (13).
Before initiation of the study, cognitive therapists (three master’s- and two doctoral-level clinicians) completed a course of weekly group supervision and treated at least two sample cases of both computer-assisted cognitive therapy and standard cognitive therapy. All sessions were audiotaped and were reviewed by an external rater. Ratings of 40 or higher on the Cognitive Therapy Scale (21) were required on at least two sequential audiotapes for qualification as a study therapist. Psychotherapy supervision was provided throughout the investigation.
Participants took no antidepressants, mood stabilizers, or antipsychotic medications for at least 2 weeks before starting the study (4 weeks for participants who had taken fluoxetine) and during the 8 weeks of treatment. Occasional use of antianxiety drugs was allowed on less than a daily basis. Alprazolam was not allowed because of possible antidepressant effects (22). Sedative-hypnotics were permitted for sleep.
Treatment manuals for computer-assisted cognitive therapy and standard cognitive therapy were developed following the general guidelines in Cognitive Therapy: Basics and Beyond(23). Therapist contact was reduced in computer-assisted cognitive therapy by using a multimedia computer program as part of therapy. In sessions 1–9, the patients initially saw the clinician and then adjourned to a separate office to work alone on the computer program. A research assistant was present during the beginning of the first session to instruct the patients on the use of the computer. Computer-assisted cognitive therapy included methods for integrating computer sessions into the overall treatment process (e.g., therapist inquiries about experiences with the computer, review of computer-assigned homework).
The software was produced first on laser disc (used in this study) and has subsequently been converted to DVD-ROM (12, 24). The learning environment, multimedia features, and initial testing of this program have been described previously (13). For this study, specific content from the program was provided at each session: 1) orientation, basic cognitive model; 2) identifying automatic thoughts and cognitive errors using thought records; 3) revising automatic thoughts, finding rational alternatives; 4) behavioral methods, scheduling activities and pleasant events; 5) further behavioral exercises, graded task assignments; 6) identifying and modifying core beliefs; 7) and 8) review and further rehearsal.
Therapists performed both types of treatment (computer-assisted cognitive therapy and standard cognitive therapy) in equal or close-to-equal proportions. Audiotaped sessions (two for each participant) of computer-assisted cognitive therapy and standard cognitive therapy were selected at random for rating on the Cognitive Therapy Scale by an external expert.
A total of 62 audiotapes were rated. Mean Cognitive Therapy Scale scores for computer-assisted cognitive therapy (mean=44.4, SD=8.4, range=23–62) and standard cognitive therapy (mean=47.3, SD=8.1, range=39–63) (t=1.29, df=60, n.s., independent samples) suggested adequate competency for both treatment modalities.
Compliance with the standard cognitive therapy and computer-assisted cognitive therapy protocols was assessed with Adherence Rating Scales, developed for this study. Adherence scores were based on completion of predetermined procedures for each therapy session and were computed as a percentage of the maximum possible points for each session. Mean percentage adherence for standard cognitive therapy sessions was 95.1% (SD=8.1, range=54.5–100), whereas mean percentage adherence for computer-assisted cognitive therapy sessions was 90.5% (SD=11.4, range=45–100). Although adherence was lower in computer-assisted cognitive therapy than standard cognitive therapy (Mann-Whitney U=255, N=62, p<0.02, two-tailed), the therapists closely followed the requirements of both treatment protocols.
The patients assigned to the wait list received no treatment during the 8-week waiting period. After completing 8 weeks of participation in the trial, they could elect to receive free treatment with computer-assisted cognitive therapy. Eleven of the 14 patients who completed the wait list period chose to receive subsequent computer-assisted cognitive therapy.
The mean age of the participants was 38.2 years (SD=9.8) for computer-assisted cognitive therapy, 41.9 (SD=9.0) for standard cognitive therapy, and 40.6 (SD=10.7) for the wait list control condition. The mean educational level was similar for all groups (computer-assisted cognitive therapy, mean=14.5 years, SD=3.2; standard cognitive therapy, mean=15.4, SD=2.7; wait list control condition, mean=15.3, SD=3.8). The percentage of female subjects in each group was computer-assisted cognitive therapy, 73.3%; standard cognitive therapy, 73.3%; and wait list control condition, 80%. Two participants dropped out of the computer-assisted cognitive therapy treatment group (by weeks 2 and 4), two from the standard cognitive therapy group (by weeks 2 and 4), and one from the wait list group (by week 3).
Examination of baseline severity variables indicated that random assignment failed to equate the treatment groups. According to one-way analysis of variance (ANOVA), the treatment groups differed significantly on the Beck Depression Inventory and the Automatic Thoughts Questionnaire. No other tendencies were evident (p values >0.15). Follow-up pairwise comparisons indicated that the standard cognitive therapy group had significantly lower severity of depression than the other two groups (t1).
Because of these differences, we examined differences in the degree of improvement across treatment with one-way ANOVA of change scores (e.g., pretreatment and week-8 scores) (25). Both intent-to-treat and completer analyses were conducted. Significant treatment effects (p<0.05) were examined with follow-up analyses by using Tukey’s honestly significant difference to control alpha (26). Also, pairwise effect sizes (27) were computed for unadjusted change scores. Complete baseline data were available except for a missing Hamilton depression scale score for one person who received computer-assisted cognitive therapy.
Changes in Depression Severity
The results are summarized in t1. Patients treated with computer-assisted cognitive therapy and standard cognitive therapy achieved significantly more improvement in depression severity than the patients on the wait list condition as assessed by both the Hamilton depression scale and the Beck Depression Inventory. Examination of between-group effect sizes for the intent-to-treat groups’ 17-item Hamilton Depression Rating Scale and Beck Depression Inventory change scores indicated large effects for the advantage of both computer-assisted cognitive therapy and standard cognitive therapy over the wait list control condition (average Cohen’s d=1.14 for computer-assisted cognitive therapy and 1.04 for standard cognitive therapy) with no evidence for differences between the computer-assisted cognitive therapy and standard cognitive therapy conditions (d=0.10). Evaluation of the time course of treatment changes (week 4 results) indicated that tendencies toward differential outcomes were evident early in the trial, but these did not reach significance.
To ensure that we adequately addressed baseline differences in severity between groups, we also examined change scores after adjusting for baseline levels of severity with analysis of covariance, treating the baseline severity score as the covariate. Homogeneity of regression assumptions was met for these analyses. Pretreatment severity scores were linked with the degree of improvement for the Hamilton depression scale but not the Beck Depression Inventory. Patients with greater Hamilton depression scale severity at baseline tended to achieve greater improvement, perhaps reflecting regression toward the mean for patients at greater levels of severity or a "floor effect" for those with mild depression at baseline. As noted, patients in the standard cognitive therapy condition had lower severity scores at baseline, as assessed by the Beck Depression Inventory. Accordingly, use of the covariate tended to increase the magnitude of the difference between the active and control treatments, particularly for the standard cognitive therapy condition. In other words, both of our approaches to data analysis provided consistent evidence for the advantages of the active treatments over the control treatment.
Maintenance of Treatment Gains
There was consistent evidence for the durability of treatment effects during the uncontrolled follow-up interval. Evaluation of depression severity at 3 and 6 months posttreatment indicated that the patients maintained their treatment gains, with mean improvement scores at least at the level of the acute treatment (week 8) results. As depicted in F1 for the Hamilton depression scale, we found a mean improvement of 11.3 points (SD=6.7) at 3 months and 11.1 points (SD=4.0) at 6 months for the computer-assisted cognitive therapy condition and 10.0 points (SD=6.2) at 3 months and 9.6 points (SD=5.5) at 6 months for the standard cognitive therapy condition. For the Beck Depression Inventory (F2), similar results were obtained; patients receiving computer-assisted cognitive therapy achieved a mean change score of 21.5 (SD=11.6) at 3 months and 21.5 (SD=9.4) at 6 months; similar longer-term benefits were evident for the standard cognitive therapy condition (a mean change in Beck Depression Inventory score of 17.6 points, SD=6.9, at 3 months and 18.9 points, SD=6.0, at 6 months). Each of these outcomes was significant for the within-group change (within-groups repeated-measures ANOVA, all p values <0.005), and the computer-assisted cognitive therapy and the standard cognitive therapy conditions did not differ significantly from each other.
Assessment of Cognitive Changes
As presented in t1, differences on the Automatic Thoughts Questionnaire were evident at week 4 (p<0.05) for the ANOVA and reflected the advantage of both active treatments over the control condition. However, there were no significant reductions on the Dysfunctional Attitude Scale, a measure of core beliefs or schemas, after only 4 weeks of treatment. Results for the Cognitive Therapy Awareness Scale indicated an early superiority for computer-assisted cognitive therapy in helping patients acquire knowledge about cognitive therapy.
At the endpoint of treatment, computer-assisted cognitive therapy, but not standard cognitive therapy, was associated with greater improvement in dysfunctional attitudes than no therapy. Computer-assisted cognitive therapy also was significantly better than no therapy in lowering scores on the Automatic Thoughts Questionnaire. Standard cognitive therapy was superior to the control condition only for the completer analysis for the Automatic Thoughts Questionnaire. Examination of the degree of change in knowledge of the cognitive approach to treatment indicated significant differences between groups, with greater increases in Cognitive Therapy Awareness Scale scores in the computer-assisted cognitive therapy group compared to both of the other conditions.
The results of this study demonstrated that both computer-assisted cognitive therapy and standard cognitive therapy produced significant reductions in major depression over 8 weeks of treatment compared to a wait list control condition. Improvement on the two primary outcome measures (the Hamilton depression scale and the Beck Depression Inventory) was evident at the end of treatment and was maintained at the 3- and 6-month follow-ups. Between-group effect sizes for the Hamilton depression scale were very large (Cohen’s d=1.14 and 1.04 for computer-assisted cognitive therapy and standard cognitive therapy, respectively). There was no evidence of differential outcome in terms of depression severity between the two active treatment conditions.
Both negative core beliefs and automatic thoughts—specific targets of cognitive therapy—were significantly reduced for patients who received computer-assisted cognitive therapy. However, only automatic thoughts were significantly lowered in standard cognitive therapy. Responses at 4 and 8 weeks on the Cognitive Therapy Awareness Scale revealed that participants in the computer-assisted cognitive therapy condition learned early in the treatment process and retained knowledge of the core concepts of cognitive therapy throughout the study.
The finding that treatment with computer-assisted cognitive therapy is associated with larger changes in dysfunctional cognitions and knowledge of cognitive therapy than standard cognitive therapy suggests that a computer adjunct may have advantages in teaching patients basic cognitive therapy methods to reduce negative thinking. Several possible explanations may warrant further study. First, the multimedia format uses video, audio, and interactive exercises to engage the patient and reinforce learning. These methods could help users better understand cognitive therapy principles and gain practice in building skills. Also, the computer program is designed to reliably deliver educational content every time it is used, whereas clinicians may choose to pay more or less attention to teaching skills for modifying dysfunctional thinking. Another possibility for the differences observed between computer-assisted cognitive therapy and standard cognitive therapy is that the computer program directly targets underlying schemas and dysfunctional attitudes for change.
The acceptability and tolerability of the active treatments, as judged by dropout rates (13%), were well within expectations for a psychosocial clinical trial (28) and approximately half that expected for antidepressant treatment trials (29). In a previous uncontrolled investigation, outpatients and inpatients reported high levels of satisfaction and acceptance of the software program, and no adverse effects were observed (13). Thus, computer-assisted cognitive therapy appears to be well accepted and tolerated by depressed patients.
Because the current sample consisted primarily of outpatients with mild to moderate depression, the extent to which the findings can be generalized to a broader or more severe population remains open to further study. Another problem in interpreting the results of this investigation is the duration of treatment. Effects were achieved with only brief treatment, consistent with our desire to study computer-assisted cognitive therapy as a strategy for aiding the efficiency of therapy. Additional studies will be required to examine the efficacy of longer forms of computer-assisted therapy. Finally, during the follow-up phase, we did not attempt to control treatment or monitor whether patients had sought additional treatment. Thus, a maintained response during this phase cannot be attributed unambiguously to the effects of acute therapy with computer-assisted cognitive therapy or standard cognitive therapy.
Several features of the research design support the integrity of the results obtained in this investigation. The study used specific inclusion and exclusion criteria, random assignment to treatment condition, evaluators blind to the condition, detailed treatment manuals, and measurement of therapist competence and adherence to treatment protocols. Although the group size was relatively small, and differences were noted between groups in baseline ratings on measures of depression, we adopted a conservative analytic strategy that assessed changes in severity from these levels. We also complemented significance testing with consideration of effect sizes.
The findings have implications for issues of treatment access and cost. Clinicians offering computer-assisted therapy were able to reduce treatment time by almost half while maintaining efficacy. This result encourages further study and application of computer-assisted methods, particularly in treatment settings in which clinician time or access to care are limited. In addition, treatment fidelity measures indicated that computer-assisted therapy was offered with a high level of precision. The ability to standardize aspects of treatment with computer-assisted cognitive therapy offers a strategy for aiding dissemination of cognitive therapy. This application may be particularly important given the low adoption rate of empirically supported psychotherapies (30, 31).
Challenges ahead for studies of computer-assisted cognitive therapy lie in examining efficacy in larger controlled trials and to ascertain effectiveness and transportability to community, primary care, and nontraditional settings. These investigations are needed to explore the promise of computerized psychotherapy: extending clinicians’ expertise to more patients at lower cost while maintaining treatment benefit.
Presented in part at the 33rd annual meeting of the Association for the Advancement of Behavior Therapy, Toronto, Nov. 11–14, 1999; the 153rd annual meeting of the American Psychiatric Association, Chicago, May 13–18, 2000; and the World Congress of Cognitive Therapy, Vancouver, B.C., Canada, July 17–21, 2001. Received June 10, 2003; revisions received March 3, 2004, and June 11, 2004; accepted July 8, 2004. From the Department of Psychiatry and Behavioral Sciences, School of Medicine, University of Louisville; the University of Wisconsin School of Medicine, Madison; New York University School of Medicine, New York; the University of Texas, Southwestern, Dallas, Tex.; Massachusetts General Hospital, Boston; and Harvard Medical School, Boston. Address correspondence and reprint requests to Dr. Wright, Norton Psychiatric Center, Department of Psychiatry and Behavioral Sciences, University of Louisville School of Medicine, P.O. Box 35070, Louisville, KY 40232; firstname.lastname@example.org (e-mail). Supported by NIMH grant MH-57470 and by a grant from the Norton Community Trust. The authors thank the following: the clinicians who served as cognitive therapists for this study, Don Kris Small, Ph.D., Peter Steiner, M.D., Troy Raffield, M.A., Virginia Evans, L.C.S.W., and G. Randolph Schrodt, Jr., M.D.; research assistants Jeffery Kuykendall and Jeffery Hudgins; study consultants A. John Rush, Jr., M.D., and Judith Beck, Ph.D.; the independent raters; and the staff of the Norton Psychiatric Center. Drs. Jesse and Andrew Wright may receive a portion of the profits, if any, from sales of software used in this investigation. All other authors will receive no royalties or other payments from software sales. A portion of the profits, if any, will be donated to the Foundation for Cognitive Therapy and Research and the Norton Foundation.
Mean 17-Item Hamilton Depression Rating Scale Scores for Depressed Patients Who Completed Computer-Assisted or Standard Cognitive Therapy or Were Put on a Wait Lista
aGroup sizes indicate the number of participants remaining in each group at each time point. At week 8, both active treatment groups improved significantly more than the wait list control group (p<0.05). There were no significant differences between the computer-assisted cognitive therapy and standard cognitive therapy groups at the 3- and 6-month follow-up assessments.
Mean Beck Depression Inventory Scores for Depressed Patients Who Completed Computer-Assisted or Standard Cognitive Therapy or Had No Therapya
aGroup sizes indicate the number of participants remaining in each group at each time point. At week 8, both active treatment groups improved significantly more than the wait list control group (p=0.001). There were no significant differences between the computer-assisted cognitive therapy and standard cognitive therapy groups at the 3- and 6-month follow-up assessments.