The following discussion summarizes the pertinent findings from these lines of research, with emphasis on studies that associate communicative behaviors with specific encounter outputs or treatment outcomes.
Negotiated treatment and the customer approach
The work of Eisenthal and Lazare (
23,
24,
25,
26,
27,
28,
29,
30) explored the applicability of the "customer approach" to the conduct of the initial psychiatric interview in a walk-in clinic. The initial evaluation was viewed as a process of negotiation between the clinician and the patient. It was assumed that patients had one or more clinical requests or perceived needs for help. The clinician's task was to empathically elicit these requests, collect other data necessary for a clinical formulation, and then enter into a negotiation in which patient and clinician simultaneously attempted to influence each other, leaving the clinician better educated to make a more valid formulation and leaving the patient more willing to consider the clinician's suggestion (
24,
25).
The negotiated consensus model described by Levinson and colleagues influenced the "customer approach" (
31). Eisenthal and Lazare used this interview model as the ideal template for the interaction between clinician and patient in an initial interview in a walk-in clinic. Eisenthal and Lazare viewed the initial psychiatric interview as a set of transactions between the clinician and the patient, which, if performed to the satisfaction of the participants, would create the context necessary for optimal treatment outputs. To ascertain the effect of this approach on clinician and patient satisfaction, the researchers developed written questionnaires for both clinicians and patients to complete before and after the encounter, to report their expectations and the results of the encounter. In addition, Eisenthal and Lazare developed the Patient Request Form (PRF), a 75-item self-report questionnaire designed to assess the degree to which patients want each of 15 categories of services. The categories were derived from interviews of 200 patients on their first visit to a walk-in clinic by asking them what type of help they wanted. The PRF was used to assist clinicians in conducting the psychiatric interview (
24,
25).
Eisenthal and Lazare reported three salient findings about shared decision making. First, clinicians saw the value for the patient and themselves of initiating the disposition discussion in an open and democratic manner. Second, clinicians associated patient satisfaction with having the recommended treatment plan match the patient's request—a match the clinicians did not connect with their own satisfaction. And third, clinicians failed to realize how important it was for the patient to engage with the clinician in a mutual sharing of power in terms of the clinician-patient relationship.
In a study of 120 patients who were referred for further treatment from an acute psychiatric service, Eisenthal and Lazare found that 41 percent of the patients kept the referral appointment (
30). Patients who did so were distinguished from those who did not by their endorsement of statements related to participation in disposition planning and by clinicians' understanding of their requests. Patients who endorsed the statement that the clinician provided a diagnostic understanding of their symptoms were no more likely to adhere to referral appointments than those who did not endorse this statement. However, diagnostic understanding was associated with patient satisfaction, feeling helped, and feeling better. In addition, adherence to referral appointments was related to a problem- or task-centered, rather than a feeling-centered, approach to the conduct of the initial interview. These findings led Eisenthal and Lazare to conclude that the customer approach could be shaped to fit a diverse patient population, resulting in greater adherence to treatment recommendations (
30).
Eisenthal and Lazare's work was limited in its the characterization of the psychiatric encounter, because it focused only on negotiated transactions and not on other aspects of the encounter that are considered valuable to the development of a trusting professional relationship—for example, the verbal and nonverbal communication cues of the clinician, and the clinician's ability to balance data acquisition for diagnostic purposes with socioemotional inquiry and support to build rapport. Eisenthal and Lazare's work was done in the 1970s, and its applicability to the psychiatric encounter in today's mental health systems has not been tested. Their work addressed only initial interviews in walk-in clinics, and therefore the generalizability of their findings to follow-up appointments, initial appointments in other settings, and specific diagnostic groups at particular times in the course of illness has not been tested.
Therapeutic alliance
The literature on therapeutic alliance addresses the therapist's ability to develop the context necessary for treatment to be beneficial. This line of inquiry has a rich history dating back to the 1950s, principally focusing on therapeutic alliance issues in psychotherapy (
32). In a review of psychotherapy research, Krupnick and Pincus (
33) highlighted the importance of identifying the active elements that make psychotherapy effective. Within these complex treatment procedures, they noted that the influence of therapist variables on treatment effectiveness and the therapeutic alliance is particularly important.
In the context of psychotherapy research, therapist variables refer to the competency of the therapist in adhering to a particular method of psychotherapy. Crits-Cristoph and colleagues (
34) reanalyzed data from ten psychotherapy outcome studies. This meta-analysis revealed that therapist effects vary considerably and at times are large and significant. In psychotherapy research, therapist variables are controlled through the use of manual-driven psychotherapeutic interventions. Videotaped observations of treatment sessions and measures of adherence to the manual are used throughout the study to monitor fidelity (
35). The term therapist variables, however, does not capture other important aspects of the therapist that potentially play a significant role in treatment: the therapist's capacity to show warmth and empathy, the therapist's ability to develop an open interpersonal dialogue, therapist demographic variables, personality characteristics that may either enhance or limit openness in the therapeutic setting, and so forth. These variables have not been thoroughly studied and may be important to the development of a therapeutic alliance.
As early as the mid-1970s, Borden (
36) identified the therapeutic alliance as an important avenue of investigation. Since then, at least six different measures have been developed to assess the role of the relationship variable on outcome. In the largest study in which the efficacy of different psychotherapeutic approaches for the acute phase of depression were tested, the Treatment of Depression Collaborative Research Program of the National Institute of Mental Health, the therapeutic alliance was found to account for more of the variance in outcome than the specific technical approach that was employed (
37). Other researchers in psychotherapy outcome studies have noted this finding as well (
38,
39,
40,
41,
42,
43,
44). Other important findings from this line of research are summarized as follows. The patient's subjective evaluation of the relationship, rather than the therapist's actual behavior, has the greatest impact on psychotherapy outcome (
32,
38). The most reliable estimates of the quality of alliance are those based on patients' reports; the next most reliable are those of outside raters (
32). Attention to the here-and-now aspects of the relationship was more likely to produce beneficial results than clinician's interpretations linking the patient's current relational crises to their past experiences (
39). Neither gender combination nor androgyny appear to have a statistically reliable impact on the quality of the alliance (
32). Finally, training aimed specifically at helping therapists develop better alliances with their patients has been less successful than anticipated (
32,
38,
45).
Although the bulk of the alliance literature focuses on the therapeutic alliance in psychotherapy, research by Frank and Gunderson (
46) and by Weiss and colleagues (
47) addressed the role of the therapeutic alliance in the pharmacological treatment of schizophrenia and chronic depression, respectively. Frank and Gunderson examined the relationship of the therapeutic alliance to the treatment course and outcome of 143 patients diagnosed with schizophrenia who were followed for two years. Assessments of alliance were obtained from the Psychotherapy Status Report, a 15-item questionnaire that therapists completed monthly (
48). The Psychotherapy Status Report included six Likert scales pertaining to the patient's in-therapy behaviors that the clinical and research literature suggested would be indicative of an alliance in any form of psychotherapy. Each scale had five levels defined by clinical descriptors. Ratings on the six scales were highly intercorrelated and were combined to form a single measure of the alliance that was termed active engagement. Treatment utilization and medication compliance patterns were obtained from therapist reports (collected monthly), patient reports (collected every six months), and medical records (reviewed on an ongoing basis). Changes in patient functioning were obtained every six months from seven instruments that were administered by trained raters who did not know the patients' identities and three self-report instruments that were administered to the patients.
Results revealed that patients who were rated as having a good alliance with their treatment provider at six months were less likely to drop out of treatment subsequently, showed a greater acceptance of pharmacotherapy, and had better functional outcomes than those who were not rated as having a good alliance. In addition, the alliance accounted for 11 percent of the variance in outcome at two years, after the association between outcome and alliance at six months was controlled for.
Frank and Gunderson's study results were limited in their generalizability to typical clinical treatment settings, because study participants were recruited from a sample involved in an ongoing research study. The study measured alliance effects on treatment outcome, based on monthly appointments—not on the longer intervals between appointments typical of many of today's psychiatric treatment settings. Additionally, because of the research design, the influence of reimbursement schemes and managed-care models of service delivery could not be ascertained. Finally, the influence of demographic differences between psychiatrist and patient on alliance development and treatment outcome was not addressed.
Weiss and colleagues (
47) studied the influence of the therapeutic alliance on the efficacy of pharmacotherapy for depression. They posited that the relationship between alliance and outcome might be more powerful in depression than schizophrenia, because the hopelessness, poor self-esteem, and self-blame that are characteristic of depression are sensitive to therapist interventions, even when these interventions are framed as education about the disorder within a pharmacotherapy context. From this position they hypothesized that even pharmacological management with medication is a type of "therapy" and that it might have many of the same process mechanisms as other psychotherapies. They further hypothesized that therapists' and patients' perceptions would correlate with outcome in the pharmacological treatment of depression.
In Weiss' study, 31 patients who met DSM-III-R criteria for major depressive disorder and had a score of 14 or greater on the 17-item Hamilton Rating Scale for Depression (HAM-D) were included. Potential participants were excluded if they had suicidal intent requiring hospitalization, organic mental disorder, substance abuse, psychosis, mania, or severe eating disorders. Patients were treated with imipramine, starting at 75 mg a day with the dose increased by 25 mg every three days as tolerated. Patients were seen weekly during initiation of medication, then every two weeks until stable, and then monthly during maintenance. Medication was gradually discontinued when patients had been in remission for at least four months. Patients were seen weekly during discontinuation of the medication and then were followed for two more monthly visits. Alliance was measured by the California Pharmacotherapy Alliance Scale, patient and therapist versions (unpublished scale, Gaston L, Marmar CR, 1991). Outcome was measured objectively by the clinician with the HAM-D and subjectively by the patient with the Beck Depression Inventory. A multiple time-series design was used to investigate the alliance-outcome association within patients over time as well as across patients.
Results revealed that the alliance in pharmacotherapy was highly correlated with outcome. Overall, 41 percent of outcome variance was explained by therapists' ratings of alliance and 25 percent by patients' ratings when alliance scores were averaged across treatment sessions.
The results of the study by Weiss and colleagues stand in marked contrast with findings on alliance and outcome in psychotherapy. Weiss and colleagues found that treatment providers' perception of the alliance best predicted outcome, and the alliance ratings they observed were lower than those seen in psychotherapy. Because the study was conducted in Canada, the study's generalizability to health care systems driven by managed care and managed competition is uncertain. Also, the study measured alliance effects for monthly appointments, not appointments that may be once every three months, a common schedule in many managed care systems. Finally, the study did not assess the influence of demographic differences between psychiatrist and patient and their effect on alliance and treatment outcome.
In conclusion, although the alliance literature addresses the association between a mental health professional-patient relationship and specific treatment outputs—such as patient satisfaction, outcome of psychotherapy and pharmacological treatment, and adherence to appointments—it is not clear whether these results are generalizable to today's psychiatric encounter within managed care and managed competition systems or various reimbursement schemes and whether the research is applicable to clinician training. Also, the influence of demographic differences between psychiatrist and patient on communicative behaviors within the psychiatric encounter and their effect on encounter outputs and treatment outcomes has yet to be explored.
Content analysis of interviews
The work of Cox and Rutter in the late 1970s and early 1980s used audiotape and videotape recordings of initial psychiatric interviews with parents of children referred to an outpatient child psychiatric clinic at Maudsley Hospital in London (
60,
61,
62,
63,
64,
65). A three-phase research strategy was used. In the first phase, appropriate measures were developed and their interrater reliability was assessed. In the second phase, a naturalistic study of 36 interviews, conducted by psychiatric residents, were used to determine the range of approaches ordinarily followed in routine clinical practice, to learn whether the psychiatrist's style was consistent over different interviews, and to identify associations between these styles and the informant's response, as reflected in the factual information given and the feelings and attitudes shown. The third phase consisted of the comparison of four different interview styles that were developed from the findings of the naturalistic study and that were close to styles recommended by influential teachers and practitioners in psychiatry.
Cox and Rutter's research revealed the following findings. Most mothers who sought help with their children's problems mentioned most of the key issues without the need for standardized questioning. The authors concluded that it is desirable to begin clinical diagnostic interviews with a lengthy period that features little in the way of detailed probing and in which informants are allowed to express their concerns in their own way (
64). In addition, it is desirable to ask certain specific questions about key issues when it is crucial to know whether or not a particular symptom or problem is present; interviewers must be sensitive and alert to factual cues and choose their probes with care and attention, so that there is a focus on the essential issues identified in the informant's spontaneous comments. Systematic questioning is not perceived as unduly intrusive or lacking in understanding (
64). Specific feelings-oriented techniques on the part of the interviewer—for example, direct requests for self-disclosure by the participant, use of interpretations and expressions of sympathy, and the use of open rather than closed questions—are valuable in eliciting feelings (
65). Finally, the gathering of good factual information is compatible with the successful eliciting of emotions and feelings; however, the two aims require rather different communicative methods (
65).
Cox and Rutter's research findings did have limitations. Results were obtained from interviews of parents, not of identified patients, so the results may not be generalizable to mentally ill patient populations. The results focused on interviewing techniques, independent of the results of the interview's association with specific outcome measures. Cox and Rutter did not address the effect of psychiatrist-patient demographic differences on research findings. Finally, their findings may be specific to the United Kingdom's socialized mental health service delivery system and therefore may not be generalizable to other systems or settings.
Harrison and Goldberg (
66), recognizing the advances in training medical students and general practitioners in interviewing skills in Great Britain and the limited research in training psychiatric residents, began a course in interviewing skills for first-year psychiatric residents at the University Hospital of South Manchester. The course ran for ten weeks; teaching was in small groups with feedback from the first author of the paper and peer trainees on a series of videotaped recordings of real patient interviews. The topics covered in the program included a detailed analysis of segments of the interview that encouraged certain types of questions and behavior from the patient and discouraged others; the important aspects of particular sections of the interview, such as the presenting complaint; and attention to overall style—controlling the interview, empathy, and so forth.
For the purposes of the study, 20 first-year psychiatric residents were asked to make three videotape recordings—one before training, an interim recording made during the interview course, and one after training. The pretraining and interim recordings were used for teaching purposes. Patients were unknown to the trainees and had a variety of presenting complaints but were not acutely disturbed. Each recording was 20 minutes long, with about ten minutes devoted to eliciting the presenting complaint and ten minutes to the mental status examination. Each separate segment of the resident's speech was defined as a specific type of utterance: compound question, leading question, closed question, multiple-choice question, neutral question, no-question, and open or directive question. Each segment of speech was also defined as a specific type of good behavior: a request for clarification, supportive remark, summarizing statement, transition statement, response to nonverbal communication, delayed verbal cues, or understanding of verbal cues. For the previously defined segments of the interview, each category of utterance and good behavior was rated on a 3-point scale from 1, not used, to 3, definitely used. The ratings were made after completion of the course by the first author, who was blind to pre- and posttraining status. Using similar, unrelated recordings, interrater reliability was performed by two raters, one blind rater who rated the study videotapes and another who did not. Agreement was between 70 and 100 percent. Rater reliability was also measured by repeating the ratings for the first ten recordings after a two-month interval. This gave a Cohen's kappa value of .85 for the utterance ratings and agreement between 70 and 100 percent for good behaviors.
After the training, the residents used more closed questions and showed an increase in the use of supportive behaviors. They were significantly better at establishing the onset of the presenting complaint, identifying precipitating factors, clarifying the main complaint and other complaints, summarizing the problems, appreciating the nature of depression and suicidal ideas, clarifying psychotic symptoms when present, using directive questions appropriately, and establishing the patient's view of their illness.
Harrison and Goldberg were able to show that the use of interaction analysis systems in psychiatric residency training is helpful in changing the types of questions posed to patients and improving the gathering of relevant historical information about the course of the presenting illness. Whether these changes positively influence specific encounter outputs and treatment outcome is left to be explored. Because of the focus on initial evaluations, the content analysis technique used in this study may not be applicable to follow-up appointments, emergency and inpatient appointments, or other psychiatric specialty evaluations. Because acutely disturbed patients were excluded from the study, the generalizability of the study's findings to acute care assessments is questionable. Also, because only 20 minutes of speech was analyzed, rather than the entire interview, important communicative behaviors within the interview process may have not been identified for analysis. Finally, the study did not address resident and patient demographic differences or personality characteristics that might influence the development of a therapeutic alliance.
In conclusion, health communications research in mental health has assessed the influence of communicative behaviors on negotiating a treatment plan, patients' following through on treatment recommendations, outcome of treatment from either a satisfaction or a symptom-reduction perspective, informing providers about the communicative utterances that help in assessment or developing a therapeutic alliance, and whether interaction analysis techniques can be used to improve psychiatric residents' interviewing skills.
Much of this research was carried out from the 1960s to the early 1990s, before the significant changes in psychopharmacology and the shift to managed care models of service delivery and capitative reimbursement schemes. Therefore, its relevance to the communicative skill set necessary for a psychiatrist practicing in today's mental health care systems is limited.
The research to date implies the existence of a universal communicative skill set that can be applied to all psychiatric services within different service settings. But is the communicative skill set necessary for the evaluation of a suicidal patient in the emergency department the same as the evaluation of a patient in a walk-in clinic? Is the communicative skill set necessary for the psychiatric assessment of a seriously and persistently mentally ill patient within a program for assertive community treatment team (PACT) the same as the assessment of the same patient in a private office setting? Is the communicative skill set that a psychiatrist uses when treating a patient with borderline personality disorder the same as the communicative skill set used when treating a person with paranoid schizophrenia?
Most clinicians would agree that the communicative skill set used in each case is dissimilar. Clinicians typically possess an array of communicative styles that have been formed through experience in evaluating patients under various conditions. The communicative style used is based on the unique characteristics of the psychiatrist and the patient, as well as the context and purpose of the evaluation. Identification of the ideal communicative skill set in every varied scenario in which psychiatric services are provided—and training residents and practitioners in those communicative skills, taking into account the practitioners' unique individual characteristics—has not been attempted.