Adverse Outcomes in Group Psychotherapy

Abstract

Group forms of therapy have been growing at a rapid rate, in part because of their documented effectiveness and economic considerations such as managed care. It is therefore becoming increasingly important to assess the psychological risks of these interventions. The author provides an overview of the published literature and conference presentations on negative effects in adult outpatient groups. Although much of the literature on adverse outcomes in group therapy focuses on single risk factors (e.g., negative leader, group process, or patient characteristics), the author argues that an interactional model should be encouraged. Means of reducing casualties are also discussed, as well as methodological issues and research directions.

The role of groups in mental health delivery is increasing in prominence, fueled both by controlled empirical studies on effectiveness and by economic considerations such as those driving the managed care system.¹ As much research attests, and recent meta-analyses demonstrate, group therapy can be a potent intervention for various age, social, educational, and diagnostic cohorts.^2–5 Similarly, the ameliorative power of support groups for people who share common crises, life transitions, or chronic medical conditions is also highly publicized.⁶ However, just as it is well documented that under specific circumstances beneficial medical treatments (e.g., X-ray and medications) are capable of doing harm, it is also generally accepted that there are iatrogenic effects in psychosocial treatments.⁷

In the group therapies, there has been a dearth of scientific studies on negative outcomes. Recent findings of iatrogenic effects in peer-group interventions with acting-out adolescents⁸ and women with breast cancer⁹ have highlighted the importance of determining factors that may jeopardize treatment safety. Concerns about the potential for harm in group interventions^6,7,10–15 have led to questions such as: Does the benefit outweigh the risk to participants? What is the extent of negative outcomes? What kinds of adverse effects occur? Which individuals are at risk to be harmed? Do certain leaders present greater risk for participants? What group factors are likely to prove damaging to participants? How can the number of casualties be reduced?

This article is designed to provide a comprehensive report on the factors contributing to adverse events in multiperson outpatient interventions with adults. Because the lines between therapy groups, growth groups, and self-help groups are not always clear, papers on all three types of interventions are included. Research on family therapy is not. The article begins with brief discussions of the historical background, incidence, definitions, and characteristics of negative outcomes. The focus then changes to the major clinical and research findings pertinent to the singular contributions that therapist, group (as an entity), and patient factors make to poor outcome. The importance of researching how interactions of these factors might affect treatment outcome is discussed. In addition, means available to group therapists for reducing adverse outcomes and recommendations for future research are presented.

NEGATIVE TREATMENT OUTCOMES

Background and Incidence

Debate over the issue of negative treatment outcomes in the individual therapies was stimulated in the 1960s by Bergin,¹⁶ who coined the term deterioration effect to describe the greater variability in experimental than control groups on criterion measures in psychotherapy outcome studies. In his review of individual psychotherapy research studies, the treatment groups contained consistently greater proportions of persons who improved, and also persons who got worse than the comparison groups. Since that time, there have been three widely cited reviews,^17–19 a book,⁷ and a research project¹⁹ devoted primarily to negative effects in individual psychotherapy. There have also been two important meta-analytic studies,^20,21 which estimated a 9% to 11% negative effect size in dyadic psychotherapy.

In the group therapies, the literature is considerably more sparse. Lieberman et al.²² conducted one of the most methodologically sophisticated studies in the small-group field. In their research, 210 Stanford University student-volunteers were assigned to 18 encounter groups. Assignments were based on a stratified random sampling of sex, class year, and previous encounter group experience. There was also a control group of 38 students. Sixteen experienced leaders, representing nine widely used group technologies, were also selected for the study. They were primarily psychiatrists and psychologists. As part of their outcome research, the investigators extensively examined group casualties. They arrived at a negative change figure of 16% that included 8% casualties. A casualty was defined as “an enduring (8 months or more), significant negative outcome, which was caused by their participation in the group.” Because of the stringent criteria employed (e.g., persons highly distressed for briefer periods of time were not considered casualties), they believe their casualty figure may be conservative.

Hartley et al.²³ provided a review article on deterioration effects in encounter groups and reported appreciable variation in casualty rates, ranging from less than 1% to almost 50%. This lack of consensus reflected discontinuities between studies in features such as criteria used for casualty; leader qualifications; member characteristics, and possible investigator bias. Hartley et al. also concluded that the differential between Bergin's¹⁷ finding of a 10% negative change figure for individual treatment (supported by later meta-analyses) and the Lieberman et al.²² 16% figure for encounter groups could be attributed to certain distinctive properties of groups rather than to unique features of sensitivity groups. However, they also believed that the extrication of the specific group characteristics causing negative effects awaited the results of future research.

Dies and Teleska¹⁴ interviewed 30 highly experienced group therapists about the incidence of their group members who actually became worse as a result of treatment. Respondents reported an average negative change incidence of 10 percent. Considering that these were expert therapists, this figure may be an underestimate of the typical incidence of negative effects in group interventions.

In the literature, researchers have used various terms and definitions for patients who get worse in therapy. This issue will be addressed in the next section.

Issues of Definition

Mays and Frank⁷ address the question of how to label therapy patient decline. They object to the term deterioration for several reasons, including that change is a complex process, in which one can simultaneously experience negative change in one sphere (e.g., become more anxious) and positive change in another (e.g., become less depressed). Strupp et al.¹⁹ advocate the term negative effect. In their usage, the term refers to a relatively lasting adverse change in a patient's condition directly attributable to the quality of the therapeutic intervention to which the patient has been exposed. It does not refer to transient effects (e.g., temporary sadness at termination) or random fluctuations due to momentary stressful life events. Mays and Frank⁷ favor the term negative outcome with no inference as to causality, whereas Dies and Teleska¹⁴ employ the same term to mean a worsening in a patient's functioning or symptoms as a result of treatment. According to Lieberman,²⁴ a judgment of casualty is typically based on a clinical finding that the person has “deteriorated in major adult role functioning.” The variations in these definitions reflect the conceptual complexities in this area and in the existing literature.

The focus in this article is on adverse outcomes that appear to be caused directly by the therapy experience (i.e., not secondary to symptom breakthroughs caused by relapse, nature of the disorder, or situational crises).

What Constitutes a Negative Therapeutic Change?

In their survey of 70 nationally respected psychotherapists (including several prominent group therapists), Strupp et al.¹⁹ found that these clinician-scholars perceived the following therapy-induced changes to constitute a negative effect: 1) exacerbation of presenting symptoms, e.g., generalization of symptoms; 2) misuse/abuse of therapy, e.g., patient substituting intellectualized insights for other obsessional thoughts; 3) undertaking unrealistic goals or tasks, e.g., pursuing goals that one is ill equipped to achieve in an attempt to please the therapist; 4) loss of trust in therapy or the therapist, e.g., patient's disillusionment prevents him or her from seeking out necessary therapy in the future; and 5) appearance of new symptoms (suicide would be an extreme example). Regarding this last point, how would the clinician or researcher know that the suicide was therapy-induced? Would a suicide note have to state it directly?

Lieberman et al.²² identified casualties by employing multiple criteria, including 1) request for emergency help during the course of group; 2) dropping out of group; 3) decreased self-esteem determined by pre–post questionnaire findings; 4) peer, leader, and third-party identification of harmed participants; 5) seeking out postgroup psychotherapy as a result of something related to their group experience; and 6) member self-report of a negative group experience.

Grunebaum²⁵ stratified negative effects in terms of their degree of harm. The stratification was based on interviews with 47 psychotherapists who claimed to have experienced a harmful personal therapy. The subjects rated themselves in terms of severity of harm on the following four-point scale with the following referents: no or mild harm (“I wasn't hurt, as I got out in time”); moderate harm (“It made me doubt myself, and I became more anxious”); serious harm (“I became much more depressed and couldn't do my work”), and severe harm (“I got so anxious and depressed that I became psychotic”).

SOURCES CONTRIBUTING TO CHANGES IN GROUP THERAPY

There are various factors that influence the success or failure of group therapy. As Roback and Smith²⁶ concluded, multiperson treatment outcomes result from a complex interaction between therapist, group, and patient factors. For example, it is widely accepted that different therapists have varying levels of comfort with different patients and pathologies. The therapist's negative bias or countertransference toward a patient (e.g., an alcoholic) may also affect the group's capacity to help that individual.²⁷ However, because much of the literature on negative effects in group therapy tends to focus on individual dimensions, I will review these single-factor contributions first, and then address their joint influences.

Therapist Effects

There is controversy over whether or not the group therapist exerts as much influence in treatment outcome as does the individual therapist. For example, in group therapy, the therapeutic strategy may emphasize the growth-promoting properties of peer interaction and relationships. Dies and Teleska¹⁴ believe that “the group treatment situation is a uniquely complicated one because the permutations of influence are more diverse than in the one-to-one context” (p. 33). On the other hand, group therapy scholars such as Corey¹³ and Harpaz¹² emphasize the contributions of the leader's personality and skill in orchestrating the dynamics that are intrinsic to a well-functioning group. This section focuses on those negative factors, namely, the contributions of therapist leadership style, selection errors, and personality factors (adjustment, countertransference) to patient decline.

Leadership Style:

Lieberman et al.²² thoroughly examined the impact of leadership style on intensive small group casualties. The charismatic leaders who were overly confronting, pressuring members for immediate and highly personal self-disclosure, and who imposed their values on the participants, often failed to recognize crumbling defenses in fragile members. There was one exception, a therapist with this leadership style who appeared to be aware of fragile members and was constrained with them. Another leadership style predictive of negative outcomes was the “laissez-faire” approach in which the leader was negligent in providing adequate structure and protection for group members. Harmful effects occurred when negatively charged member-to-member feedback took place without protective norms. In his review of leadership in short-term groups, Dies²⁸ also concluded that negative leadership style (e.g., continued high-intensity negative statements by the therapist) was associated with increasing group tension, lowering of patient self-esteem, and an escalation of the risk for adverse events. Further, Dies and Teleska¹⁴ found that too much ambiguity about group goals and procedures at the beginning of the group's life appears to impede group development as well as feed patient interpersonal distortions, interpersonal fears, and subjective distress. The attribution of adverse events to a given leadership style (as noted with the therapist who was the exception in the Lieberman research) is likely to be confounded by therapist personality attributes and clinical skills.

Selection Errors:

Selection errors frequently result in a mismatch between therapeutic modality and a patient's severity of preexisting psychopathology. This mismatch can lead to additional therapist mistakes such as probing more deeply than the patient can tolerate.¹⁹ Gilmore²⁹ discussed a case of a female patient (diagnosed as schizoid) inappropriately switched from individual therapy to psychodynamically oriented group therapy. The author noted that “the group's more complex set of interactional and interpersonal rules exceeded her ability to imitate appropriate functioning, overwhelmed her with the possibility of rejection, and highlighted her intense object needs while defeating her ability to reduce others to thing-like objects” (pp. 7–8). The patient was traumatized by her experience and terminated her participation in the group. In some forms of group intervention, little attention is paid to screening issues. For instance, in the self-help movement, there is often no screening of the appropriateness of potential members for the group by a professional mental health clinician. In some instances the group process can reinforce undetected psychopathology in a group member, and such individuals, reciprocally, can undermine a supportive group process.²⁸

Personality Factors:

Therapist negative countertransference³⁰ (e.g., direct expressions of hostility, lack of respect, and sexual acting out) is associated with harmful consequences for group members. Lothstein³¹ found that group therapists in his study had “disliked the patients” who later became what the investigator referred to as “therapist-induced dropouts.” Holahan³² discussed the countertransference phenomenon as it applies to the group therapist running multiple groups. The author examines how his failure to understand his countertransference towards his “least favorite group” contributed to patient dropout. Specifically, he referred group therapy candidates “with the most capacity to care for one another” to his other groups, and referred members “with the highest potential for narcissistic uninvolvement” to the group that had frequently verbally attacked him during a period of painful personal loss. The composition of the latter group (his “least favored child”) left the members “less prepared to deal with the inevitable storms involved in intensive dynamic group therapy” (p. 4). Holahan believed that if group therapy failure events in clinical practice were carefully examined, clinicians would discover their own contributions to the unproductive events. According to Holahan, “The failure results from a systematic series of decisions on the part of the therapist, consciously and unconsciously motivated, decisions which may be fueled by therapist narcissism and therapist pain” (p. 2).

Therapist personality maladjustment is another major factor associated with harmful outcomes. In the study by Lieberman et al.,²² the verbally aggressive, intrusive, and overly confident group leaders who had the most casualties (and were unable to detect those participants' increasing distress) appear to be severely narcissistic, defective in empathy, and unconcerned about their group members' emotional needs. On the basis of interview and questionnaire data, Grunebaum²⁵ placed the clinicians described as destructive by their psychotherapist-patients into the following categories: distant and technically rigid (e.g., had difficulty relating to their patients in ordinary human ways—were like insecure technicians following fixed rules of procedures); emotionally seductive (e.g., fostered intense feelings in patients without helping them understand their emotions); and highly unethical (e.g., having severely crossed therapeutic boundaries by engaging in sexual activity with patients). This finding of distant and uninvolved therapists, overly intense therapy, and severe boundary violations causing harm to certain patients is similar to the findings of Lieberman et al.²²

In a symposium on iatrogenic issues in group therapy held at the 1993 American Group Psychotherapy Association meetings,¹¹ two highly respected therapists openly discussed the role of iatrogenic conditions in their personal group therapy experience approximately a decade earlier. Both presenters implicated the group therapist's abuse of power, which translated into a negative domination and exploitation of the patients. One of the speakers reported feeling powerless to confront her therapist's excessively domineering behavior; instead, by conforming to it, she permitted herself to be controlled by the group leader. As that individual so eloquently said, “My true self remained in hiding while my false self was praised and reinforced over and over again.” The other speaker discussed the grossly exploitative and damaging behaviors of his group therapist, who acted out sexually with group members, encouraged sex between co-members, and humiliated members by having them wear self-demeaning signs during group sessions. The therapist's rationalization for such unethical behavior was that he was attacking their narcissistic defenses! The phenomenon of surrender by group members to a psychologically disturbed therapist (i.e., their participation in the pathological events), with resultant negative outcomes, needs to be further researched.

Dies³³ commented that therapeutic experience and competence would be expected to affect treatment outcome, especially in light of the inherent complexity of group interventions and the difficulty of mastering them. However, I could find no studies that examined the relationship between the group therapist's experience level, the committing of therapeutic mistakes (e.g., making technical errors or underestimating the degree of patient psychopathology), and poor outcomes.

Group Process Effects

A group is often more than the sum of its parts. At times, however, it may be less than the sum of its parts. Ideally, therapeutic groups develop a work culture under the skillful direction of a leader knowledgeable not only in the areas of psychopathology and psychodiagnostics, but also in group dynamics and interpersonal communication. That is, characteristics of the group itself become critical in treatment outcomes. Dynamic properties of therapeutic groups include factors such as intragroup cohesion, group norms, group roles, group pressure, conformity, communication structure, social comparison, and self-disclosure. For purposes of this paper, we are interested in identifying the group processes most relevant to negative effects.

Lieberman et al.²² found that attack or other rejection of a member by the group (or leader) was among the primary mechanisms of injury. That is, problematic factors in the group process, such as highly critical interpersonal feedback about one's personal shortcomings, are particularly potent in the absence of a cohesive group climate. It is helpful for a person to learn, in a safe and supportive group context, about the inappropriateness of his or her interpersonal behavior. However, it is quite damaging for the recipient to be attacked viciously by co-members in a group in which there is little solidarity. In this same study, “feedback overload” was considered another mechanism with potential for antitherapeutic effects. That is, disparaging feedback that is delivered in an overly confrontational fashion by co-members can lead to considerable confusion rather than a helpful new self-perspective. Dies and Teleska¹⁴ believe that scapegoats and persons assuming deviant group roles are more likely to experience significant negative effects (i.e., be harmed by the group's wrath) if attacked at specific developmental stages of the group. MacKenzie³⁴ also believes that circumstances that are likely to increase the risk of negative effects can be predicted on the basis of group development. For example, before a group has become consolidated into a strong, supportive unit, a patient's disclosure of “highly charged material (e.g., incest) might lead to rejection because the group may find that information too overwhelming” (p. 219). However, it appears that little, if any, research has studied possible linkage between stage of group development and negative outcome. Lieberman et al.²² also implicate group reactions to a member's catharsis as being a critical factor for a negative experience. That is, if the participant's strong emotional expression is met by group silence or rebuff, rather than understanding and support, the individual may well have a destructive experience.

Groups, unlike individual therapy, are relatively public and permeable by their nature. Yet there is relatively little research on the short-term and long-term negative effects of co-member confidentiality breaches (i.e., norm violations) on the targeted individual.³⁵ For example, an unauthorized disclosure about a member's extramarital affair can lead to possible divorce, job loss, and other severe repercussions. Galinsky and Schopler⁶ found that confidentiality violation was a major concern of group leaders for cancer support groups. Some facilitators believe that members might make statements out of their despair, or other motivations, that would put them at risk if these disclosures were leaked outside the group circle.

Individual Member Effects

Several patient characteristics have been linked to early group dropout and/or patient decline. In several studies and reviews,^{14,22,36–38} premorbid level of psychological disturbance was found to be an important correlate of negative change. Patients diagnosed as severely narcissistic, borderline, or schizoid appeared to be most at risk. Such individuals tend to have difficulty forming an alliance with co-members and are thus at high risk to assume deviant group roles. Another patient characteristic associated with poor outcome is unrealistic expectations. Persons who changed negatively in the Lieberman et al. project²² did not anticipate pain or discomfort as part of the therapy process. Mohr¹⁸ views patients' expectations about therapy as “emblematic of their approach to the world. It is not unlikely that this group of people walk blindly through the world stumbling from one catastrophe to the next.… Psychotherapy is just one more calamity” (p. 13).

In addition to level of psychopathology and unrealistic therapy expectations, four other pregroup risk factors for persons vulnerable to a negative outcome have been identified.^22–24,36 These pregroup characteristics included 1) severe self-esteem problems; 2) the combination of poorly developed interpersonal skills and high interpersonal sensitivity; 3) a tendency to assume deviant group roles; and 4) being conflicted about self-disclosure and intimacy. Similar patient characteristics were identified as risk factors in persons prematurely dropping out of group treatments.^26,39 Perhaps those who experience treatment failure are susceptible to develop an oppressive sense of being outside the group culture, feel painfully alone to face the entire group, and/or poorly tolerate negative feedback.³⁴ Using factor analysis, one could determine the relative percentage of the variance in negative outcomes accounted for by risk factors, alone and in combination with each other.

Goeltz⁴⁰ discusses negative therapeutic reactions from a psychodynamic perspective. Unfortunately, it is not possible to tie in the specific hypotheses offered in his article with the actual findings on patient characteristics reported earlier in this section. On the basis of personality style, Goeltz expects certain persons to be at high risk to suffer a poor therapeutic outcome, in part because of the negative countertransference they engender in the therapist. The self-critical, masochistic type of patient with an attachment to suffering may verbally attack the therapist in order to evoke retaliatory rage (“sadistic countertransference”). Fears of fusion and dependency are two other dynamic issues for patients at risk for negative outcome, according to Goeltz. To defend against feelings of terror over being incorporated or merged with the therapist, such patients may prove competitive with the leader in order to emphasize their separateness from him. The countertransference in this situation may be characterized by the therapist's wish to engulf the patient. In providing a therapeutic climate for the types of patients described by Goeltz, the therapist needs a high degree of theoretical expertise, understanding, and tolerance for frustration. Similarly, in the group therapies, it is important for the therapist to recognize that some participants are highly anxious about what the group might do to them, even as they hope for what the group might do for them.

RESEARCH CONSIDERATIONS

An Interactional Model

The isolation of factors associated with patient decline may be useful in the initial stages of inquiry, but a richer understanding comes from considering these important components simultaneously. Bach⁴¹ observed, “Certain patients who have the most intensive conflicts in one group find it relatively easy to communicate and participate in another group” (p. 301). In essence, Bach is referring to the compatibility, or fit, of patient and group. Life-threatening effects have been reported for patients with chronic lung disease whose breathing problems worsened when participating in a group treatment with a psychodynamic format.⁴² The patients experienced respiratory distress during periods of intense emotional expression. The format of the group was changed to a didactic emphasis, with favorable results. Just as a patient may not “fit” comfortably in a specific group, or may be mismatched with a specific group approach, there are also obvious instances of poor patient–therapist matches. Similarly, it is highly likely that there are therapist–group miscombinations, as illustrated in the Holahan³² paper. Clearly, the three dimensions (therapist, group, and member) operate simultaneously in mediating treatment outcome. In their book on the empirical bases for the group ecosystem, Fuhriman and Burlingame⁴ discuss studies that investigated interaction effects between combinations of two factors (e.g., the impact of members' locus of control on treatment effectiveness in directive and nondirective group therapies). However, the studies cited did not directly address negative outcomes.

Methodological Issues

Systematic research into the interaction of factors primarily responsible for negative group therapy outcomes is a formidable challenge.⁴³ Kazdin⁴⁴ provides important design recommendations for research on adverse outcomes, including instrumentation, statistical power issues, and multiple outcome effects. For example, he notes that the conclusions reached about positive or negative treatment effects might vary at different points in time. He provided an example of an outcome study in which positive short-term treatment effects were followed by long-term deleterious effects. In discussing instrumentation, Kazdin discusses how negative change scores may not reflect a counterpart in the person's daily performance or functioning. That is, change in a negative direction from pretreatment to posttreatment may be a function of artifacts in the measuring instrument rather than actual changes in behavior. Mohr¹⁸ also addresses the difficulty of accurately determining the meaning of negative change on assessment instruments. He focuses on the importance of clinical significance. According to Mohr, whereas a patient's negative pre–post difference score may reach statistical significance, “it could also reflect a less defensive posture on the part of the patient in reporting difficulties and feelings, or it may be a last-ditch effort to stave off the end of a time-limited therapy” (p. 20). That is, a negative pre–post therapy change in an outcome measure may not necessarily signify decline. However, one might use that same argument to claim that a positive pre–post difference could reflect a patient's intent to please the therapist! It is important that assessment instruments in therapy outcome research have validity measures to detect positive and negative impression management. Both Mohr¹⁸ and Kazdin⁴⁴ recommend the use of multiple outcome measures, since dimensions of change often covary (i.e., decrease in one symptom may covary with increases and decreases in other symptoms). Mohr also notes the confounding of relapse and negative outcome, as well as ethical issues (e.g., how far are negative responders allowed to decline before they are removed from the protocol?) and political issues (impact of negative psychotherapy findings on an institution's reputation as well as that of its clinicians).

Research Directions

Some suggested directions²⁶ for future research on negative events include investigations into whether or not the incidence of negative outcomes varies according to specific group applications (e.g., cognitive, psychodynamic, gestalt), length of treatment (short-term versus long-term), and open-ended versus closed group formats. Are negative impacts of group therapy qualitatively different from those in dyadic therapy? The relationship between family environmental variables (e.g., family dysfunction) of participants and negative group process has also received minimal attention.⁴⁵ Studies on improving the quality of treatment for high-risk patients by focusing on the therapy process itself would be a major contribution.⁷ For example, the empirically validated cognitive-behavioral, life skills approach of Linehan,⁴⁶ which combines individual and group formats, each with specific goals and techniques, appears promising with high-risk, severely ill borderline patients. However, I am not aware of any studies examining adverse events and outcome with these empirically validated procedures. In fact, they may have given us a false sense of security. An important new area of research in the individual therapies is patient profiling.⁴⁷ This approach includes a statistical methodology based on patient intake data for predicting how the patient should respond (positive and negative outcome) to treatment. The potential for patient profiling in the group therapies should also be explored.

RECOMMENDATIONS FOR REDUCING CASUALTIES

Despite the paucity of empirical evidence regarding many of the factors suspected to be important in affecting quality of outcome from group interaction, some preventive measures can be at least tentatively recommended for group therapists. For instance, several clinician-scholars^{6,7,10,13,16,23,26,27,47} have offered practical suggestions for reducing negative treatment outcomes emerging from what we do know, and what we believe, to be important in this area. These recommendations are blended together here for economical presentation of the material.

With respect to group members, proposals include early and accurate identification of high-risk patients (e.g., screening for persons likely to become “group deviants” and subject to rejection by co-members). Greater attention needs to be paid to group preparation. Members should have realistic expectations concerning the process of therapy and improvement in order to prevent patient–group or patient–therapist conflict resulting from dissimilar expectations. If the individual is not suitable for a specific group for reasons of personal-interpersonal style, or if mutual expectations cannot be achieved, then the individual should be referred to a pretherapy training group (to help him or her better understand group process), a more suitable group format, or alternative forms of care. High-risk individuals may also need to be in concurrent individual and group therapy.

Leader-related proposals include having therapists arrange for peer support. High-risk patients have typically alienated most people in their lives, and therapists need open discussion with a respected colleague about real relationship problems or countertransference issues with such individuals. In terms of leader training, therapists and other group facilitators need to be specifically trained to prevent or reduce nonconstructive confrontation and to identify patients being harmed by it. Further, trainee self-awareness groups that permit feedback about problematic styles of leadership should be part of professional training programs. Certification of group therapists has already taken place and will help prospective group members identify well-trained group providers. Rutan and Stone²⁷ believe that “therapy groups evoke powerful feelings in all who sit in them, including the therapist” (p. 151). They suggest that the therapist fantasize about patients in his current group whom he would like to exclude. Such approaches might help therapists learn about their countertransference and might be proactive before negative events occur. Corey¹³ reminds us that it is the group leader “who is responsible for minimizing the inevitable psychological risks associated with group activity” (p. 35).

DISCUSSION

In terms of risk–benefit analysis, there is ample scientific evidence of the effectiveness of group treatments. Many people have been profoundly helped by the ameliorative power of group forces. However, the current report is not about efficacy; it is about psychological safety. The singling out of group treatments does not imply that these are the only psychosocial interventions requiring closer scrutiny. Mohr¹⁸ believes that the field of psychotherapy has shied away from looking at negative outcomes and seldom reports them. He asserts: “To the extent that the field avoids examining when psychotherapy fails, the field succeeds only in limiting its own potential” (p. 24). As is often found in psychotherapy outcome studies, negative responders are “buried” in the outcome variance¹⁸ or are simply not reported.⁸

Because of the dearth of empirical research on negative outcomes in traditional therapy groups, I have also here presented findings from growth groups and self-help groups. Although these three group approaches differ in focus, there is overlap in their constituencies and mobilization of group forces for purposes of personal change.^24,48 Information from the qualitative methodology literature (e.g., case reports) was also included. Although retrospective reports by patients and therapists can legitimately be criticized for potential bias and contamination,²⁸ they are often useful for generating research hypotheses, particularly when negative outcome is currently such an understudied area.

The current review suggests that the three major dynamics (leader, group, help-seeker) typically associated with facilitating positive outcome are also implicated in negative outcomes. Findings suggest that problematic aspects of the therapist get transmitted through “negativistic” intervention styles (e.g., rejecting comments and behaviors), misapplication of technical skills (e.g., interpretations that the individual does not have the resources to integrate in a meaningful way), and harmful relationships with the group or with particular individuals. Negative group process characterized by low cohesion and hurtful social interactions (e.g., scapegoating) also contributes to psychological injury in vulnerable individuals. Further, each group member's personality style and adjustment play important roles in the resulting group dynamics and subsequent treatment outcome. Persons with severe character pathology (narcissistic, borderline, or schizoid) appear at particular risk for poor outcome, underscoring the delicacy of group process for fragile individuals. However, as Lieberman and colleagues²² note, “The phenomenon of psychological injury is a complex and varied one” (p. 193). Their intensive study of group participants led them to conclude that some persons had experiences similar to those of the casualties, yet were left untouched or even benefited from the experience; other persons suffered a significant decline that was viewed as group-specific (i.e., the individual likely would have had a more positive outcome in another group); and some individuals' vulnerabilities would likely have resulted in a hurtful experience regardless of which group they had participated in. When discussing negative effects, we are also left with the conundrum of whether therapy was the catalyst for the patient's decline or if the regression would have occurred without treatment.

The next steps in negative outcome research will be to determine the type and frequency of casualties across group applications (e.g., comparing psychodynamic, interpersonal, and cognitive-behavioral group formats for treating depression) and to gain a richer understanding of how the interactional flow between the individual and co-members (including the leader) produce negative outcomes. Admittedly, group and therapist effects are not easily separated (an example would be therapist influence on group norms). However, as Rosenbaum et al.⁴⁸ note, “The challenge to develop this kind of knowledge grows greater with the rapid increase in the bewildering variety of helping experiences that use group approaches” (p. 720). The present article is intended to serve both as a reference and a stimulant for such research in our quest to better understand and reduce iatrogenic effects in the psychosocial therapies, as is being done for pharmacological treatments.