Solution-Focused Brief Therapy: A Review of the Outcome Research

Abstract

Solution-focused brief therapy (SFBT) is a new and increasingly used therapeutic approach that focuses on helping clients construct solutions rather than solve problems. The approach evolved in a clinical context amid many anecdotal reports of success from both therapists and clients, but it has not been subjected to controlled empirical testing until very recently. In this article we critically review all of the controlled outcome studies of SFBT to date (N = 15) to assess the extent to which SFBT has received empirical support. Five studies were well-controlled and all showed positive outcomes—four found SFBT to be better than no treatment or standard institutional services, and one found SFBT to be comparable to a known intervention: Interpersonal Psychotherapy for Depression (IPT). Findings from the remaining 10 studies, which we consider moderately or poorly controlled, were consistent with a hypothesis of SFBT effectiveness. We conclude that the 15 studies provide preliminary support for the efficacy of SFBT but do not permit a definitive conclusion. Our critique highlights areas where methodology in future studies can be strengthened to provide more conclusive evidence of SFBT efficacy.

In less than two decades, solution-focused brief therapy (SFBT) has grown from a little-known and unconventional therapeutic approach to one that is now widely used in the United States and, increasingly, in other countries. SFBT is used in family service and mental health settings, in public social services and child welfare, in prisons and residential treatment centers, in schools and hospitals (Miller, Hubble, & Duncan, 1996). Enthusiastic practitioners report successful outcomes and high client satisfaction using SFBT. Insurers and governmental funders have increasingly embraced SFBT because it is short-term and therefore relatively inexpensive.

But widespread use and anecdotal reports of success do not provide an adequate basis for the ongoing use of SFBT, or any therapeutic approach. What is needed is objective, empirical evidence of the effectiveness of SFBT—that clients are better off in demonstrable and meaningful ways as a result of intervention. Accordingly, we decided to conduct a comprehensive review of the available outcome research to see to what extent there is empirical support for the effectiveness of SFBT. We begin by providing a brief description of SFBT and early attempts to document SFBT outcomes. Then we critically review controlled studies of SFBT outcomes that have appeared in the literature through 1999—fifteen studies in all. Based on our review, we discuss the extent to which SFBT has received empirical support and conclude with recommendations for the kind of additional research that is needed to establish SFBT clearly as an empirically supported treatment.

SFBT origins

SFBT evolved out of the clinical practice of Steve de Shazer, Insoo Kim Berg, and colleagues at the Brief Family Therapy Center in Milwaukee, Wisconsin, in the early 1980s (de Shazer, 1982, 1985, 1988; de Shazer, Berg, Lipchik, et al., 1986). As the name suggests, SFBT is defined by its emphasis on constructing solutions rather than resolving problems. The main therapeutic task is helping the client to imagine how he or she would like things to be different and what it will take to make that happen. Little attention is paid to diagnosis, history taking, or exploration of the problem. Solution-focused therapists assume clients want to change, have the capacity to envision change, and are doing their best to make change happen. Further, solution-focused therapists assume that the solution, or at least part of it, is probably already happening (Weiner-Davis, de Shazer, & Gingerich, 1987). Treatment is brief, usually lasting less than six sessions.

Over the years, de Shazer, Berg, and colleagues developed a number of specific techniques to aid in solution-focused intervention. The best known of these is the miracle question, which asks the client to pretend that a miracle has happened and imagine a solution to the problem (DeJong & Berg, 1998; de Shazer, 1988). A second technique routinely used is the scaling question, which asks the client to rate on a 10-point scale how things are today. Both of these techniques are used to aid in the construction of the solution and the search for parts of the solution that may already be happening. SFBT typically includes a “consulting break” toward the end of the session in which the therapist constructs a message that includes compliments for the client and a homework task.

Early followup studies

As the solution-focused model was evolving, the team at the Brief Family Therapy Center conducted followup surveys of clients to determine whether clients were benefiting from the new approach. Treatment outcome was measured by asking clients at 6–18 months followup to indicate if they had met their goals for therapy or felt that significant progress had been made. In the first such study, de Shazer (1985) reported an 82% success rate on followup of 28 clients. The next year, de Shazer et al. (1986) reported a 72% success rate with a 25% sample of 1,600 cases. Subsequent studies have reported similar results (DeJong & Hopwood, 1996; Kiser, 1988). The data from these studies compare favorably with those reported earlier by Weakland, Fisch, Watzlawick, and Bodin (1974) who followed up brief therapy clients in Palo Alto.

Several clinician-researchers outside the Milwaukee group have also conducted followup studies of SFBT (Lee, 1997; Macdonald, 1997; Morrison, Olivos, Dominguez, et al., 1993; Schorr, 1997). For the most part, these studies also used subjective outcome measures and found similar but somewhat smaller success rates. Schorr (1997), however, employed a pre-post design with the State-Trait Anger Inventory as the outcome measure, and found that after 8 group sessions the percentage of members scoring in the clinical range had declined from 67% to 40%.

While these followup studies provided important early feedback on SFBT outcomes, their lack of experimental control does not permit causal inferences to be made about the effectiveness of SFBT. Recently, however, controlled studies of SFBT outcomes have begun to appear in the literature, and they are of primary interest in our review.

Review methodology

We decided to review all of the controlled studies of SFBT client outcomes appearing in the English literature up to and including 1999. By controlled studies, we mean studies that employed some degree of experimental control, that is, used a comparison group or single-case repeated-measures design. By client outcomes, we mean client behavior or functioning—we excluded studies that reported only client satisfaction. Finally, we limited our review to studies that reported end-of-treatment or later outcomes. This ruled out studies that examined intermediate therapy outcomes, such as the impact of the Formula First Session Task on the second session (Adams, Piercy, & Jurich, 1991). Finally, we excluded ethnographic and change process studies (Beyebach, Morejon, Palenzuela, & Rodriguez-Arias, 1996), since, by definition, they did not assess end-of-treatment outcomes.

We identified the domain of potential outcome studies by first searching the PsychLIT, Social Work Abstracts, PsychInfo, Medline and Dissertation Abstracts bibliographic databases, selecting studies that used the terms “solution focused” or “solution-oriented” and “outcome research” in their titles or abstracts. We augmented this by searching the World Wide Web on the same terms. We also reviewed a listing of published outcome studies compiled by Alasdair Macdonald, Research Coordinator of the European Brief Therapy Association (Macdonald, 1998a). Finally, we consulted several reviews of SFBT research (DeJong & Hopwood, 1996; Franklin & Jordan, 1999), as well as the bibliographies of each of the outcome studies we included in our review.

To be considered a study of solution-focused brief therapy, the intervention had to be identified by the study's author(s) as solution-focused or solution-oriented as developed by de Shazer and colleagues at the Milwaukee Brief Family Therapy Center (de Shazer & Berg, 1997; Macdonald, 1998b). Operationally, this meant the intervention had to include one or more of the following core components: (1) a search for pre-session change, (2) goal-setting, (3) use of the miracle question, (4) use of scaling questions, (5) a search for exceptions, (6) a consulting break, and (7) a message including compliments and task.

The final selection of studies was made after we independently reviewed each potential study and then together agreed that the study

• • 
  implemented solution-focused brief therapy,
• • 
  employed some form of experimental control,
• • 
  assessed client behavior or functioning, and
• • 
  assessed end-of-treatment outcomes.

Our search located 15 controlled studies of SFBT outcomes appearing in the literature through 1999. We were guided in our critique of the 15 studies by standards for assessing empirical support for psychological treatments developed by the American Psychological Association (Task Force on Promotion and Dissemination of Psychological Procedures, 1995) and modified by Chambless and Hollon (1996). In brief, these standards require studies to (1) use a randomized group design or acceptable single-case design; (2) focus on a specific, well-defined disorder; (3) compare the experimental treatment with a standard reference treatment, a placebo or, less desirably, no treatment; (4) use treatment manuals and procedures for monitoring treatment adherence; (5) use outcome measures with demonstrated reliability and validity; and (6) use a sample large enough to detect group differences reliably.

Based upon these standards we have divided the studies into three groups according to the degree of experimental control they employed: (1) five studies met 5 to 6 standards and we consider them well-controlled, (2) four studies met 4 of the standards, and we consider them moderately-controlled, (3) six studies met 3 or fewer standards and we consider them poorly-controlled. (See Tables 1, 2, and 3, below.)

Table 1. 

Table 2. 

Table 3. 

Well-Controlled studies

Depression in college students

Sundstrom (1993) was the first to design a randomized experimental study of SFBT using standardized outcome measures. She compared a single session of SFBT to a single session of Interpersonal Psychotherapy for Depression (IPT) for the treatment of depressed college students. The sample was comprised of 40 female undergraduate psychology students at a Midwestern university who scored in the mild to moderately depressed range (10–29) on the Beck Depression Inventory. Thirty-four percent of the sample met diagnostic criteria for Major Depression according to the Inventory to Diagnose Depression.

Subjects were randomly assigned to the experimental or control group, and then given the battery of outcome measures. Measures included the Beck Depression Inventory (BDI), the Depression Adjective Checklists (DACL), the Rosenberg Self-Esteem Scale (SES), and the Counselor Rating Form–Short Form (CRF-S). Treatment consisted of one 90-minute counseling session—SFBT for the treatment group and IPT for the comparison group. Sessions were conducted by 21 female counselors who were licensed social workers, licensed psychologists, psychology interns, or advanced psychology graduate students. To assure adherence to treatment protocols, clinicians participated in separate 2-hour training sessions for each condition, and all counseling sessions were videotaped and rated by research assistants who were blind to the treatment condition. A followup interview was conducted a week to 10 days after treatment, at which time subjects completed the BDI, DACL, SES, and CRF-S.

MANOVA analysis of pre- to post-intervention BDI and DACL scores showed both treatment conditions produced significant positive change, and that neither treatment produced significantly better outcomes than the other. SES scores revealed no change across time for either treatment condition. Lack of significant differences between treatments in CRF-S scores indicated counselor characteristics did not contribute differentially to treatment outcome.

Sundstrom's study satisfies most of the standards for demonstrating empirical support. She used a randomized control group design, studied a well-defined sample, compared SFBT with a standard reference treatment, used treatment manuals and monitored adherence, and used standardized outcome measures. It is not clear how the subjects (college students) were selected for the study; however, Sundstrom established that subjects were mild to moderately depressed and that many (one-third) met diagnostic criteria for major depression. Although Sundstrom found no significant differences between treatments, and no trend favoring one treatment over the other, the number of subjects (20 per group) was too small to conclude reliably that the treatments were equivalent. Sundstrom's study demonstrated that single-session SFBT was effective in reducing depressed mood, and suggests that SFBT outcomes may be comparable to IPT, an empirically validated treatment for depression (Chambless, Sanderson, Shoham, et al., 1996). Uncertainty about the representativeness of the study sample makes it difficult to know to what clinical populations Sundstrom's findings may generalize (see Table 1).

Parenting skills

Zimmerman, Jacobsen, MacIntyre, and Watson (1996) evaluated the effects of a solution-focused parenting group on parenting skills and perceived family strengths. Parents experiencing difficulties with their adolescents' behavior were recruited through a newspaper advertisement and respondents were randomly assigned to an experimental (N = 30) or wait-list control (N = 12) group. All subjects completed the Parenting Skills Inventory (PSI), an 86-item self-report questionnaire with 7 subscales designed to measure parenting skills, and the Family Strengths Assessments (FSA), a 12-item measure of family happiness.

Five experimental parenting groups, each composed of 6–8 parents, met weekly for six 30-minute sessions. Graduate students in the Marriage and Family Therapy Program at Colorado State University facilitated the groups under faculty supervision. Direct comparisons of the post-test scores of the SFBT and wait-list groups revealed statistically significant (t-test) differences for the total PSI score and several PSI subscales: Role Image, Rapport, Communication, and Limit Setting. No statistically significant differences were found between groups for the FSA. Analysis of pre-post data for the SFBT group revealed statistically significant change for the Role Image, Objectivity, Communication, and Limit Setting subscales of the PSI, and the FSA total score. In contrast, pre-post comparisons for the wait-list group showed significant change for only one subscale of the PSI: Role Support. The authors note that even though the SFBT group showed significant pre-post improvement on FSA scores, it was not significantly different from the wait-list group at post-test.

The use of a randomized pre-post design, treatment manual, and standardized outcome measures suggests that subjects benefited from the SFBT intervention. However, there are a number of considerations that qualify this conclusion. The small size (apparently due to high attrition) of the wait-list comparison group (N = 12) makes it impossible to conclude reliably that the wait-list subjects did not also improve during the study, but, unfortunately, the authors do not address the issue of trend in the wait-list group. If by chance the subjects in the wait-list group did improve during the study, the efficacy of SFBT would be cast into doubt since, presumably, some third factor caused change in both groups.

A strength of the Zimmerman et al. (1996) study is that it targeted a specific population (parents experiencing adolescent conflict); however, subjects were self-selected, casting doubt on what populations the findings can be generalized to. Finally, comparison of SFBT subjects with wait-list (“no treatment”) subjects does not control for attention effects that may go along with treatment, resulting in what Chambless and Hollon (1998) call a possibly efficacious but not specific treatment. In other words, although subjects appeared to benefit from the intervention, it cannot be determined that the benefit was due specifically to the SFBT intervention as opposed to the nonspecific effects that presumably accompany any intervention.

Rehabilitation of orthopedic patients

Cockburn, Thomas, and Cockburn (1997) evaluated the impact of SFBT on psychosocial adjustment and return to work for patients with orthopedic injuries. The study sample was comprised of 48 patients and their spouses, referred by an orthopedic surgeon to a rehab program designed to prepare patients for work re-entry. Subjects had to be first-time recipients of a worker's compensation claim, married to a spouse who is employed full-time, not currently on prescription drugs, and without other medical problems that could compromise the study. Seventy-three percent of the final sample had orthopedic injuries of the spine or upper extremities.

Subjects were randomly assigned to one of four groups, following a Solomon Four Groups design. The intervention for treatment groups 1 and 3 consisted of 6 weekly one-hour sessions of SFBT plus the standard rehab program. Treatment was implemented by the first author and followed a standard protocol (Jack Cockburn, personal communication, April 18, 1999). Control groups 2 and 4 received only the standard rehab program. Pre-test data were collected from treatment group 1 and control group 2 using the Family Crisis Oriented Personal Evaluation Scales (F-COPES). Post-test data were collected from all 4 groups using the F-COPES and the Psychosocial Adjustment to Illness Scale–Self-Report (PAIS-SR). Subjects' spouses also completed the PAIS-SR at post-test. Because pre-testing was shown to have a consistent effect across treatments, analyses were based on ANOVAs for the post-test data only.

Analysis of F-COPES data indicated significant between-groups differences on all 3 subscales used in the study. The two treatment groups were 9–13 points higher on Acquiring Social Support, 9–14 points higher on Reframing, 5–7 points higher on Mobilizing Family, and 4–5 points higher on Seeking Spiritual Support than were the two control groups. Between-group differences on the PAIS-SR were also statistically significant. Couples in the treatment groups were about 4 points higher on the Health Care scale, 4 points higher on Domestic Environment, 9 points higher on Psychological Distress, and 5 points higher on Social Environment. There was no significant difference between groups on the Vocational Environment scores. The authors concluded that patients in the SFBT groups had significantly better psychosocial adjustment and social supports than patients in the control group. Within 7 days after completion of treatment, 68% of subjects in the treatment groups had returned to work as compared to only 4% of subjects in the control groups. By 30 days after treatment, 92% of the SFBT patients had returned to work as compared to 47% of control group patients.

This study employed a rigorous, randomized design, screened subjects according to well-defined eligibility criteria, used a treatment protocol, and used standardized outcome measures along with a measure of ultimate outcome (Rosen & Proctor, 1978)—return to work. Differences in return to work rates were of sufficient magnitude to have obvious clinical significance. Although the sample size was small (25 treatment subjects, 23 control subjects), it was sufficient to demonstrate that the SFBT group was significantly improved as compared to the standard treatment group. Since this study compared SFBT with the standard rehabilitation protocol (which presumably did not include individual counseling with patients), there is no control for attention effects; thus, we do not know to what degree SFBT effects were specific. However, the data clearly demonstrate that SFBT plus the standard rehabilitation care was superior to standard care alone. This finding was most evident in increased return-to-work rates, the ultimate goal of rehabilitation programs such as this one.

Recividism in a prison population

Lindforss and Magnusson (1997) studied the effectiveness of a SFBT network intervention in reducing recidivism for prisoners incarcerated at Hageby Prison in Stockholm. This population consists of serious criminals with high recidivism rates, disciplinary problems, and long histories of drug abuse and contact with correctional and social welfare agencies. Prisoners who had 2–10 months left to serve and were willing to participate in the study were randomly assigned to a SFBT treatment group (N = 30) or a control group (N = 30). Treatment was provided by a team consisting of a project leader and two family therapists who were in private practice in Stockholm. Treatment lasted for 1–12 sessions, with an average of five sessions.

Recidivism, defined as committing an offense subsequent to release, which resulted in probation or reincarceration, was used as the outcome measure. Data were taken from central prison and probation records at 12 and 16 months after subjects were released from prison. At 12 months after release, prisoners in the treatment group had a recidivism rate of 53% compared to 76% for the control group. At 16 months, recidivism rates increased to 60% for the experimental group and 86% for the control group, and differences remained statistically significant. In addition, the seriousness of recidivist offenses and length of resulting sentences were less for the SFBT group than the control group. Finally, the authors note that the prisoners in the control group “incurred an expenditure of 2.7 million Swedish crowns more in prison costs than the experimental group during the followup year” (Lindforss & Magnusson, 1997, p. 102).

This study is notable for its use of a well-defined and difficult to treat population, random assignment to groups, and use of an ultimate outcome measure—recidivism. Although arrest and adjudication data are subject to biases in criminal justice processing, they have obvious validity as a measure of ultimate outcome. Nevertheless, other standardized measures of behavior outcomes would have been useful in assessing treatment outcomes. Further, the study is limited by lack of a treatment manual or procedures for monitoring treatment implementation. Sample size, while not large, was adequate for detecting differences between the treatment and control groups. While it is unclear exactly what the SFBT network intervention consisted of in this study, it is evident that prisoners who received the treatment benefited in terms of lower recidivism rates. This is also one of very few studies that addressed directly or indirectly the issue of cost-benefit of SFBT intervention.

Antisocial adolescent offenders

Seagram (1997) evaluated the efficacy of SFBT for improving attitudes and behaviors, and reducing antisocial thinking and behavior in adolescent offenders in a secure facility for youthful offenders. The study sample consisted of 40 youths who were rank-ordered according to sentence and then alternately assigned to the treatment (N = 21) or control (N = 19) group to insure comparability on seriousness of offence. Participants had to have a diagnosis of psychosis and a history of refusal to take medications to be eligible for the study. Eighty-five percent of the sample had a history of violent behavior, 90% were repeat offenders, and 65% were currently incarcerated for a violent crime.

All subjects attended a group orientation session and three individual assessment sessions prior to the treatment participants beginning 10 weekly SFBT sessions. SFBT was offered in addition to the standard services provided by the institution. The author, a doctoral candidate at York University in Ontario, conducted the orientation, assessments, and treatment, with each SFBT session lasting 45–60 minutes. An external reviewer rated first and last sessions to insure adherence to the SFBT model.

Outcome measures included the Jesness Behavior Checklist Recidivism Scales, Achenbach Youth Self-Report, Carlson Psychological Survey, Coopersmith Self-Esteem Inventory, and a Solution-Focused Questionnaire developed by Seagram. The Jesness was completed by the youth, his correctional officer, and a teacher; the other measures were all self-report.

Scores on the Solution-Focused Questionnaire indicated that the treatment group made more progress in solving problems and had higher confidence in their ability to maintain changes than did the control group. Carlson Psychological Survey data indicated that the treatment group had significantly more optimism for the future, greater empathy, fewer antisocial tendencies, and less chemical abuse. Treatment group subjects showed significantly less difficulty with concentration (Achenbach). The Jessness and Coopersmith measures failed to show any significant between-group differences. Teacher ratings on the Jesness and Teacher Report Form showed trends favoring the SFBT subjects, but differences did not reach significance. Within a 6-month followup period, 4 (20%) members of the treatment group vs. 8 (42%) members of the control group had re-offended (run away or were moved from open to secure custody).

Seagram's study meets many of the criteria for demonstrating empirical support. A matching design was used (in lieu of random assignment) to assess whether SFBT improves on the outcomes of standard institutional care; the study sample was well-defined; although a treatment manual was not used, treatment adherence in the first and last sessions was rated by an outside observer. Outcome measures included several widely used objective measures of behavior change. Although Seagram found some significant between-groups differences, and several trends in the expected direction, overall SFBT outcomes appeared modest. Finally, comparing SFBT with standard institutional (noncounseling) care did not allow for an estimation of the specific effect of SFBT.

Moderately-controlled studies

Counseling high school students

Littrell, Malia, and Vanderwood (1995) used a randomized post-test only design to examine the effects of three variants of single-session brief therapy on alleviating academic and personal concerns and increasing goal achievement of students at a large, urban high school. Sixty-one male and female students in grades 9–12 who sought appointments to discuss problems with their school guidance counselors were randomly assigned to one of three groups: problem-focused with a task, problem-focused without a task, and solution-focused with a task. All groups implemented two or more of the following four steps common to brief treatment: (1) define a problem, (2) identify previously attempted solutions, (3) set a specific goal, and (4) assign an intervention task. The two problem-focused groups (with and without task) implemented steps 1–4 or 1–3, respectively, whereas the solution-focused group (SFG) implemented only steps 3–4. Counselors received training in brief counseling methods and consulted treatment protocol sheets developed by the researchers.

Outcome data were collected at brief followup sessions 2 and 6 weeks after the initial counseling session. Students and counselors collaboratively rated student progress in three areas using 7-point Likert-type scales: alleviating concerns; moving toward goals; and decreasing of intensity of undesirable feelings, thoughts, and actions. All three models of therapy showed statistically significant improvement across all three areas of change between the 2-week and 6-week followups. However, no significant between-treatment effects were found (see Table 2).

Although this study reported positive outcomes, its internal validity is compromised by several methodological limitations. Most importantly, standardized outcome measures were not used and pre-treatment measures were not administered. Although subjective ratings by counselors and clients have obvious face validity, they are subject to social desirability biases that may render them invalid as measures of actual client functioning. This, and the lack of pre-test data, prevents us from concluding that change occurred due to treatment. Further, relatively minor differences between the three therapy models make the study more a comparison of various components of brief therapy than a controlled test of SFBT itself.

Solution-focused school groups

LaFountain and Garner (1996) investigated the impact of solution-focused groups (SFG) on school age children and school counselors. School counselors were recruited for a training program in SFG as an alternative approach to managing large caseloads with fewer resources. The final sample included 57 counselors randomly assigned to treatment and control groups who served a total of 311 elementary, middle, and high school students. Experimental group counselors attended a full-day SFG training workshop and were then asked to select 4–8 students from their caseloads who met criteria for inclusion in SFG. Treatment consisted of 8 weekly SFG sessions. Control group counselors (who were told the purpose of the study) did not provide any type of group counseling to their students. They were asked instead to identify potential students for SFG intervention and administer pre- and post-measures only to those students.

The Index of Personality Characteristics (IPC), a 75-item questionnaire that measures child functioning, was administered prior to intervention and again 8 weeks later (at the completion of the intervention for SFG students). Modest but statistically significant between-group differences were found on 3 IPC subscales: Nonacademic, Perception of Self, and Acting In. According to the authors, these differences suggest that students in the experimental group had higher selfesteem in nonacademic arenas; more positive attitudes and feelings about themselves; and more appropriate ways of coping with emotions.

Because counselors (who were randomly assigned) knew the purpose of the study and selected the students to participate in the study, it is impossible to determine if student characteristics relevant to treatment outcome were randomly distributed between the groups. Further, this study occurred in a naturalistic setting and was not directed toward a specific diagnostic group, making it difficult to determine exactly what to what population the results might generalize. It did, however, use a treatment manual and standardized outcome measure. Although the data indicate that SFG subjects benefited, it is difficult to know to what extent the results are due to the intervention as opposed to some other factor such as selective assignment to groups or counselor expectations.

SFBT training for supervisors

Triantafillou (1997) conducted a pilot study of the impact of solution-focused training for mental health supervisors on client outcomes. The setting for the study was a private children's residential treatment agency in Ontario, Canada. Typical client problems included depression, stress, hyperactivity, and frequent acute episodes of aggressive and antisocial behaviors. The entire supervisory and management staff of the agency and one 7-person team of direct care workers attended training in SFBT. Training consisted of four 3-hour weekly sessions that introduced the solution-focused model and taught specific solution-focused techniques for suicidality, anxiety and arousal disorders, motivational issues, anger management, crisis intervention, and supervision. Supervisors followed a set of guidelines as they implemented the SFBT approach.

The experimental client group was comprised of 5 children served by the team of direct care workers who, along with their supervisor, had participated in the solution-focused training. The children were 10–14 years old, and 4 of the 5 had a minimum of 15 months residency in the agency. The control group included 7 children from the agency's residential program matched for age, length of residency, and presenting problems. The outcome measure used was the frequency of serious incidents, and client use of psychotropic drugs as reported in the agency's records. Data were gathered at 16 weeks following SFBT supervision training. Agency records showed that the treatment and control groups had similar numbers of serious incidents of psychotropic drug use during the 9 months prior to the intervention.

During the 16 weeks following SFBT supervision training, the number of serious client incidents in the treatment group had decreased by 65.5% as compared to a 10% decrease for the control group. During the same period, two of the clients in the treatment group were able to discontinue their psychotropic drugs completely whereas 66% of the control group clients increased their dosages.

Although random assignment was not used in this study, matching was employed in an effort to equate the experimental and comparison groups. It is difficult to determine, however, exactly what the intervention consisted of in this study since the supervisors of both the experimental and comparison groups were trained in SFBT. It appears the difference was that the childcare staff for the experimental group was trained in SFBT whereas the staff for the comparison group was not. Standardized outcome measures were not employed, but the use of serious incidents (for example, use of physical restraint) lends validity to the assessment of outcome. However, since serious incidents involve staff responses to child behavior, they are subject to bias in a nonblind study—it is possible, for example, that staff trained in SFBT reported fewer incidents rather than the children actually having fewer such incidents. Doubts about the validity of the outcome measure, along with the small sample and uncertainty about what actually constituted the intervention make the findings difficult to interpret.

Couples therapy

Zimmerman, Prest, and Wetzel (1997) investigated the effectiveness of a solution-focused couples therapy group for improving marital satisfaction. Married couples who wanted to improve their relationship were targeted through a newspaper advertisement and respondents were assigned to the treatment group (N = 23 couples). The comparison group (N = 13 couples), who did not identify themselves as dissatisfied, were recruited through flyers posted in university married student quarters and a childcare center. Six treatment groups, each consisting of three to five couples, met weekly for 6 weeks. Groups were facilitated by male and female co-therapists who were students or recent graduates of the Marriage and Family Therapy Program at Colorado State University. Sessions were audiotaped and monitored by faculty supervisors to insure adherence to the SFBT model.

All subjects completed the Dyadic Adjustment Scale (DAS) and the Marital Status Inventory (MSI) at pretest. The DAS was used to assess marital adjustment at pretest and post-test. The DAS provides a total score for global dyadic adjustment as well as 4 subscale scores: dyadic consensus, dyadic satisfaction, affectional expression, and dyadic cohesion. The MSI was used at pretest to verify that the two groups were not significantly different in terms of likelihood to divorce.

DAS scores for the treatment group revealed statistically significant improvement on the total score and all 4 subscale scores. At the end of treatment, post-test scores for the treatment group had improved to the point where they approached the pretest scores of the nondistressed control group. In addition, couples' self-reports of change were congruent with these findings.

Strengths of this study include the controlled setting (university teaching clinic) in which it took place, use of a treatment protocol and verification of treatment adherence, use of standardized measures at pre- and post-test, and careful statistical analysis of differences between groups at pretest as well as pre- to post-test change. Instead of using a randomized design, however, this study recruited treatment subjects from a self-identified distressed population and control subjects from a nondistressed population. Pre-treatment comparisons indicated the groups were similarly committed to their relationships but that the treatment group was more distressed, suggesting that the two groups were somewhat representative of their presumed populations. The fact that the distressed group's scores at post-test approached those of the nondistressed group lends clinical validity to the results.

Poorly-controlled studies

Problem drinking

Polk (1996) used a single subject AB design to investigate the effectiveness of SFBT on problem-drinking behavior. The client was a 36-year-old male who had a 10-year history of problem drinking and poor work attendance. Baseline data were reconstructed from archival records and client historical report. Findings indicated that the abstinence from alcohol and work attendance both increased over the course of treatment. At baseline, the client had been abstinent one day per week; by the end of treatment he was abstinent three days per week. Work attendance during baseline had been as low as two days per week, but increased to 4–6 days per week during treatment.

Because reliable measures were used in the context of an AB design, we can be fairly confident that the client's behavior in this study actually changed; however, we don't know if it was statistically or clinically significant. Further, because AB designs are open to many threats to internal validity, it is not possible to conclude that the SFBT intervention is what caused the behavior change here. A more rigorous design (ABAB or multiple-baseline) and several replications across subjects (at least 3) would be needed to draw firm conclusions about the efficacy of SFBT (Chambless et al., 1996). See Table 3.

Family environment

Eakes, Walsh, Markowski, et al. (1997) studied the impact of SFBT on families with a member diagnosed as schizophrenic. The sample included 10 patients and their families being served by a community mental health center. The first 5 families to volunteer were assigned to the treatment group and the next 5 were assigned to the control group. Treatment and control groups met every other week with a psychiatric nurse for standard aftercare consisting of 20-minute medical checks. Immediately following each of these checks, for five times the treatment group participated in a SFBT session facilitated by the study authors and a psychiatric nurse supervisee. The SFBT group showed significant increases on several dimensions of the Family Environment Scale: Expressiveness, Active-recreational Orientation, and Incongruence, whereas the control group showed significant decreases.

The generalizability value of this study is severely limited by the small sample and nonrandom assignment to groups. The finding that treatment families improved significantly suggests SFBT had a positive outcome, but lack of controls for influences, such as therapist (authors of the study) expectancies, precludes such a conclusion. Further, the deterioration of the control group during the study is puzzling, raising questions as to the nature of the comparison intervention—what it did or did not consist of.

Outpatient family counseling

Franklin, Corcoran, Nowicki, and Streeter (1997) report data from three single-case AB studies of families experiencing parent-adolescent conflict. All three cases were selected post hoc from the agency's files to illustrate the clinical utility of using single-subject designs and self-anchored scales. Baselines were reconstructed, and intervention consisted of approximately four weekly outpatient family sessions using SFBT. Outcome measures were self-anchored rating scales developed collaboratively by therapist and client. Case 3 also used the Family Adaptability and Cohesion Evaluation Scales III (FACES III) and the Family Satisfaction Inventory (FSI). The self-rated outcome measures showed positive behavior change for all three cases, which the authors judged to be statistically significant. Case 3 showed large pre-post improvements on the cohesion scale of the FACES III, and the FSI.

These three studies are subject to all of the validity threats of AB studies and, in addition, are subject to selection bias as well. The fact that subjects were selected post hoc from agency records also raises serious questions about the representativeness of these cases and their generalizability to a typical clinical population. Subjective client ratings of outcome are also subject to reliability and validity problems. Thus, it is difficult to know exactly how much client change occurred during these studies, whether it was due to the intervention, and to what extent these results might generalize to other clients and other settings.

SFBT: public social services

Sundman (1997) examined how solution-focused ideas might change social worker-client relationships and empower clients in a welfare agency in Helsinki, Finland. The 11 social workers who agreed to participate in the study received a 20-hour workshop in solution-focused methods and ongoing supervision during the research project. Each social worker personally decided how to implement the solution-focused ideas presented in the training, however. The nonrandom control group was comprised of 14 social workers from comparable agencies in Helsinki. Three hundred eighty-two clients were selected randomly from the caseloads of the experimental and control group social workers. Outcomes were measured using a questionnaire in which social workers reported on their clients' problems and goals, progress toward goals, and helpfulness of the social work relationship. Although both groups appeared to have improved, no statistically significant differences in goal achievement were found.

This study is of interest because it is the only study to have taken place in a large public social service setting. However, it lacks most of the criteria needed to establish empirical support: most importantly, a well-defined and consistently implemented intervention, and a clearly defined and unbiased measure of outcome.

SFBT: school consultation

Geil (1998) used single-case AB designs to compare the impact of behavioral consultation, solution-focused consultation, and no consultation on the behavior of students in an elementary school. School psychologists were asked to identify teachers requesting consultation about students exhibiting externalizing behaviors in the classroom. The final sample consisted of 8 teacher-student pairs. Psychologists were assigned to treatment condition based on time and availability for training; all received 6 hours of training over a 6-week period as related to their assigned group. Psychologists met with teachers twice each week to consult regarding the identified students, and teachers were responsible for implementing all interventions. Three teacher-student pairs were assigned to the solution-focused consultation, two were assigned to behavioral consultation, and three to no consultation. Trained observers coded child behaviors during the sessions, with teachers using the Code for Instructional and Student Academic Response (CISAR), an interval-paced time sampling observation system that measures academic response, task management, and competing response behaviors.

Geil (1998) found that variability within and between baseline and intervention phases in each of the 8 cases precluded unequivocal conclusions regarding the relationship between consultation and reduction in externalizing student behaviors. In only two cases, one student in the behavioral group and one in the solution-focused group, did client change seem significant. Although the AB design precludes unequivocal conclusions about the impact of intervention, internal validity of this series of studies was strengthened by replication across 8 subjects, use of a detailed treatment manual, use of a standardized behavioral outcome measure, and careful analysis of the client change data.

SFBT: outpatient mental health

Lambert, Okiishi, Finch, and Johnson (1998) investigated the efficiency of SFBT for treating typical outpatient mental health problems, such as mood disorders, anxiety, adjustment disorders, and substance abuse in adults. The experimental group consisted of 27 consecutive adult patients of an experienced psychologist and researcher (one of the authors) committed to solution-focused methods. The comparison group was comprised of 45 patients who were treated with time-unlimited eclectic psychotherapy by trainees at a university outpatient clinic (Kadera, Lambert, & Andrews, 1996). Client outcomes were measured using the Outcome Questionnaire (OQ), a 45-item client self-report instrument designed to measure patient change over short time periods by assessing symptomatic distress, interpersonal problems, and social role adjustment. The OQ was administered prior to each session. Patients whose OQ scores decreased at least 15 points (the reliable change index) and fell below 63 (the clinical cut-off score) were considered to be recovered.

Thirty-six percent of the 22 SFBT patients whose initial OQ scores were above 63 were recovered after 2 sessions of SFBT, and 46% were recovered after 7 sessions. This compared with 2% of the comparison group recovered after 2 sessions of time-unlimited eclectic treatment, and 18% recovered after 7 sessions. Twenty-six sessions were required before 46% of the comparison sample had recovered. The authors noted that factors other than treatment approach may have influenced the results of their study, most notably the significantly greater professional experience of the psychologist as compared to the trainees, and the lack of time-limits in the comparison study.

This is the only study of the 15 we reviewed that directly examined the efficiency of SFBT, one reason many practitioners give for using SFBT in their practice. However, nonrandom assignment to groups, wide disparities in therapist experience, and no apparent emphasis in the comparison group on treatment efficiency render the results inconclusive. An important contribution of this study, however, is its use of empirically derived criteria for assessing the clinical significance of client outcomes.

Discussion

Although we located 15 controlled studies of SFBT outcomes, methodological problems in many of the studies make it difficult to interpret their findings. Accordingly, the following summarization of results is based primarily on the five studies we consider well-controlled. All five of these studies used samples of 40 or more, all employed random (or matched) assignment to groups, most used some means to insure consistency in treatment implementation, and most employed established measures of outcome.

All five of the well-controlled studies reported significant benefit from SFBT—four (Cockburn, Thomas, & Cockburn, 1997; Lindforss & Magnusson, 1997; Seagram, 1997; Zimmerman, Jacobsen, MacIntyre, & Watson, 1996) found SFBT to be significantly better than no treatment or standard institutional services. Since these studies did not compare SFBT with another psychotherapeutic intervention, we are not able to conclude that the observed outcomes were due specifically to the SFBT intervention as opposed to general attention effects. The other well-controlled study (Sundstrom, 1993) compared SFBT with a known treatment (IPT) and found SFBT produced equivalent outcomes (no significant differences were found). Since none of the five studies met all of the stringent criteria for efficacy studies, and all five studied different populations (that is, there were no replications by independent investigators), we cannot conclude that SFBT has been shown to be efficacious. We do, however, believe that these five studies provide initial support for the efficacy of SFBT. The remaining ten studies contain methodological limitations that preclude drawing firm conclusions, but we note that their findings are consistent with the general conclusion of SFBT effectiveness. One of those studies (Lambert, Okiishi, Finch, & Johnson, 1998) suggests that SFBT may be an efficient intervention (that is, successful outcome can be achieved with minimal intervention).

Although all of the studies we reviewed purported to show positive outcomes, we have reservations about their representativeness; specifically, we have no way to determine if these 15 studies are in fact representative of all SFBT outcome studies carried out to date. It is possible that investigators who found negative results did not submit their work for publication, or that journal editors chose not to publish studies with null results. We note, however, that several of the studies reviewed here made it into the literature even though they used weak designs or had modest outcomes.

However, we believe these 15 studies indicate that progress is being made in subjecting SFBT to empirical test. Early case reports and followup surveys have been followed more recently by controlled outcome studies, albeit imperfect in many respects. But the capability and resources to conduct rigorous efficacy studies take time to develop, and it appears that this evolution is taking place in SFBT outcome research. We urge investigators to continue this development—to improve the rigor of outcome studies—so that reliable conclusions about the efficacy of SFBT can be reached. To that end, we offer several suggestions based on our appraisal of the research to date.

Perhaps the area of greatest need is the specification and proceduralization of SFBT itself, and the consistent use of detailed treatment manuals and treatment adherence measures. Our analysis indicates that only two studies implemented all 7 components of SFBT; five studies implemented 4 or fewer components. Further, the modality of SFBT implementation varied widely among the studies, including individual and family interventions, supervisory and group interventions, and a network intervention used in one study. While this suggests SFBT may be widely applicable, efficacy can only be established if the intervention is clearly and specifically proceduralized so that it can be implemented consistently by different investigators across research sites. This process of proceduralization will likely require more specification of the intervention than has heretofore taken place.

A second limitation of most of the studies reviewed here is that they were conducted by advocates of SFBT. In many cases it appears that the SFBT intervention was implemented by the authors themselves. This is particularly problematic when the comparison group consists of no treatment or ordinary institutional services, because such designs leave uncontrolled the effects of therapist expectancies and therapist allegiance, which one would expect to be reasonably strong and influential when treatment is implemented by enthusiastic advocates. Future efficacy studies will need to compare SFBT with other empirically validated interventions where therapist allegiance is equally balanced between treatments.

Other considerations include specification of the study sample, selection of the comparison group, and adequate sample size. Many of the existing studies did not use screening criteria to insure that subjects met the stated characteristics. In most psychotherapy studies this is accomplished by use of DSM diagnostic criteria (American Psychiatric Association, 1994) or standardized screening instruments. Given the aversion of solution-focused therapists to diagnosing clients, based on their belief that diagnosis is not relevant and may even be harmful (Miller, 1997; Miller & de Shazer, 1998), it is not surprising that many of these studies did not employ strict selection criteria. Apart from the theoretical and practical arguments for and against diagnosing, however, it is clear that the type of presenting problem affects the choice of outcome measures to be used in a study. Clearly there must be some means of determining which measures will be relevant to the changes subjects are expected to make as a result of treatment. Using conventional diagnostic groupings would accomplish this objective, and make it possible to compare SFBT outcomes with other intervention approaches.

For the reasons noted above, we believe it is desirable for the comparison group to be another psychotherapeutic intervention, preferably an empirically validated one. This makes it possible to determine if SFBT is as good as a known intervention. However, it is necessary for the samples in such studies to be large enough (40–50 subjects per group) to have sufficient power to reliably conclude there are no differences, that is, that one treatment is equivalent to another in outcome (Cohen, 1988). This is a very difficult and expensive requirement to meet, however, and one that is rarely met in psychotherapy research (Kazdin & Bass, 1989).

As we noted at the outset, our review of SFBT outcome research is based on accepted efficacy criteria. The conventional view is that an intervention must be shown to be efficacious in highly controlled settings before studying its effectiveness in ordinary practice settings (Chambless & Hollon, 1998; Koss, Butcher, & Strupp, 1986). Some investigators have taken issue with this view, noting that, in the real world, effectiveness studies often precede efficacy studies (Clarke, 1995; Fraser, Nelson, & Rivard, 1997; Hoagwood, Hibbs, Brent, & Jensen, 1995). A more appropriate view may be to see them along a flexible and bidirectional continuum whereby studies can include a combination of features of both efficacy and effectiveness studies. Although the 15 studies reported here fall short of what is needed to establish efficacy, they provide some evidence of treatment effectiveness—evidence that SFBT works in typical practice settings.

Finally, although the stringent criteria set forth by the American Psychological Association for efficacy studies have been influential, there is growing recognition that highly controlled efficacy studies may not be feasible or even possible in service delivery settings, and if they are, they may be sufficiently lacking in clinical validity as to be of questionable use (Goldfried & Wolfe, 1998; Persons & Silberschatz, 1998; Pinsof & Wynne, 2000). Of particular issue are the efficacy requirements that a study sample be homogeneous and strictly screened, and that an intervention be defined in its purest form and rigidly implemented across all subjects. Critics maintain that real-world clients cross diagnostic groupings, and that interventions are often tailored to the individual client and integrated with other intervention approaches. Others have questioned the necessity of using randomized clinical trials, as compared with correlational designs, for estimating treatment effects (Benson & Hartz, 2000; Concato, Shah, & Horwitz, 2000). It appears that the issue of what type of research (efficacy vs. effectiveness) is possible to carry out, and what type will produce findings of use to clinicians, is coming under increasing question. Regardless of the resolution of this issue, we believe it is possible to approximate more closely efficacy standards in SFBT outcome research, and we urge investigators to work toward that end.

Conclusion

Although the current studies fall short of what is needed to establish the efficacy of SFBT, they do provide preliminary support for the idea that SFBT may be beneficial to clients. The wide variety of settings and populations studied and the multiplicity of modalities used suggest that SFBT may be useful in a broad range of applications, however, this tentative conclusion awaits more careful study.

Having examined the evolution of outcome research over the past two decades, it is evident that SFBT is moving from an “open trial” phase of investigation to an “efficacy” phase. Several investigators seemed to acknowledge as much by characterizing their research as “pilot studies” or “preliminary investigations.” Clearly, the studies reviewed here are moving in the direction of efficacy research. They, along with numerous case reports and earlier followup studies, provide a foundation for conducting more rigorously controlled investigations that can provide more conclusive evidence of SFBT efficacy.