Abstract
Why explore the concept of the therapeutic alliance in case management?
The primary model of service delivery in integrated community mental health services within Australia (case management) is endorsed at the national level [1, 2]. Consequently, this model has been widely adopted by those responsible for service provision, and much money, time and effort invested in its implementation at the National, State and Territory government level. However, the assumption that case management was a superior method of service delivery that led to improved patient outcomes in community mental health services had no objective basis in fact [3–5]. While the assumption may have face validity, there is no hard evidence to date to support the overall adoption of this method of service delivery for all patients in community mental health services.
Outcomes and case management: the evidence
Historically, case management had its roots in deinstitutionalization; the primary measures of effectiveness have been focused on service outcomes such as reduced hospitalization and bed days. It was widely recognized that the more intensive forms of case management, particularly those based on the principles of the Programme for Assertive Community Treatment (PACT) [6] were effective in reducing hospital use for patients with serious and persistent mental illness [5]. However, outcome research has been less than definitive regarding other outcomes such as quality-of-life, symptomatology and functioning, treatment compliance, and the successful rehabilitation of the patient to optimal community living [3–5].
Limited research into case manager processes in some studies have tended to focus on objective measures such as case manager inputs, that is, number of contacts and/or time invested and, to a lesser extent, the type of activity involved. These are factors primarily linked with the intensity of contact associated with such models rather than any other aspect of the model or process. However, research into the actual process by which case managers facilitate positive outcomes, such as those identified, and successful rehabilitation of the patient to community living is sparse [5, 7, 8]. Furthermore, the evidence which exists for efficacy of certain case management models in selected patient populations does not identify the critical factor/s involved in achieving the limited outcomes demonstrated. Although the importance of the patient–case manager relationship to positive outcome was implicit in some models of case management [9, 10] it remained a largely unexamined area with no attempt made to define, measure or follow the relationship or its effect on outcome in case management.
What next?
Currently the concept of patient empowerment through ‘partnership or alliance’ is central to the aims of service delivery in attempts to reduce the impact of mental illness on the individual, their families and society; and aid recovery of the individual in the community [1, 11, 12]. Providers of mental health services seek delivery strategies that facilitate these aims, guide resource allocation and maximize effectiveness. Therefore, it is timely to examine this neglected area of the patient–case manager relationship and its effect on outcome for the mentally ill patient.
Exploration of the construct of the therapeutic relationship within case management offers an avenue by which the interactive process between patients and case managers may be ‘unpacked’ and analysed. The complexity of questions regarding the patient–case manager relationship includes: is the relationship more important in facilitating a therapeutic milieu than a given model or professional discipline orientated mode of service delivery? (The emphasis for case management teams to date.) What effect do the pre-existing characteristics of the participants have on the relationship? For example, is it the personal characteristics or the professional training of the case manager that makes a difference, or neither, or both? Does age, gender, experience or postgraduate training make a difference? Similarly for the patient; does diagnosis, symptomatology, age, gender, experience of the system, hospitalization, drug/alcohol use, exposure to trauma, education, employment or social skills influence the ability of the patient to form constructive relationships with health care professionals?
This paper seeks to stimulate debate on the role of the relationship in case management in a community mental health service and its effect on outcomes such as quality of-life, symptomatology and functioning, treatment compliance and the successful rehabilitation of the patient to optimal community living. Theoretical issues associated with definition and measurement of the relationship and the evidence for its effect on patient outcome are examined. Bordin's model [13–15], which uses the term ‘Working Alliance’ to define the relationship, is proposed as the one with most utility for case management in mental health. Research implications of the role and function of the alliance in case management are discussed, together with the potential relevance of this type of research to service delivery. Finally, a proposed research study into the Working Alliance in a multidisciplinary case management service in community mental health is briefly outlined.
Background
Prerequisites to the practical application of the therapeutic alliance in case management requires that it is definable, measurable and related to improved patient outcomes.
Definition
No conclusive definition of the therapeutic relationship exists. Psychodynamic theory suggests that the relationship centres on issues of transference based on the Freudian origins of the concept, while contemporary thinking views the relationship as a reality-based experience. The latter is focused on the ‘here and now’ and the ability of the patient and clinician to work together in the interest of problem resolution.
Historical development
The therapeutic relationship had its genesis in the early 1900s when Freud used the term to describe the specialized relationship that existed between the ‘healer and patient’ [16]. Freud's initial idea of the relationship revolved around the concept of positive transference as the mechanism by which a positive therapeutic alliance, and hence successful therapeutic outcome, was facilitated. Subsequently, Freud broadened his early view of transference to include a contemporary reality-based perspective, from which Greenson first developed the concept of the Working Alliance [16, 17].
Empirical research into the relationship began in 1976 [18] with early models still firmly embedded in the psychoanalytical framework and the concept of positive transference. At this time, the concept was further developed as a dynamic rather than static process that was responsive to differing stages of therapy [17]. These historical foundations shaped the evolutionary development of the relationship and have continued to influence its contemporary expression.
Contemporary development
Currently, four principal theoretical constructs (see Table 1) based on the work of Greenson (1965), Rogers (1951–1957), Strong (1968) and Bordin (1976) have framed the definition, terminology and measurement of the therapeutic relationship [17, 19]. Together, these constructs represent a spectrum from a neo-Freudian, classical psychoanalytical view, to a reality-based perspective of the therapeutic relationship. The theorists outlined in Table 1 have all attempted to identify a key therapeutic variable that could be applicable across a range of therapies [15] and have used differing terms to describe the relationship, including the ‘therapeutic’ and ‘helping’ relationship or alliance, the ‘therapeutic bond’ and the Working Alliance [19, 20]. For the researcher seeking to expand the concept outside that of classical psychotherapy, this impedes examination of the literature as a cohesive whole, as issues of definition and measurement are intrinsic to the theoretical perspective of the relationship.
Four contemporary theoretical perspectives of the therapeutic relationship
The collaborative interdependent nature of Bordin's alliance as outlined above, was a notable departure from other theorists such as Rogers and Strong, who while acknowledging the patient-centred focus of therapy, saw the patients as passive recipients in the process, with therapeutic gains essentially being the result of therapist based activity. It is important to note that the theorists, primarily from the psychotherapy field, viewed the patient as someone actively seeking change [14]. While this may be so in general psychotherapy or counselling fields, it is not necessarily a given, at least initially, in patients with mental health illness, some of whom are in involuntary treatment.
Model of choice for case management
Bordin's model [13, 14], offers the most utility for the case management field. It is the most inclusive in terms of cross-disciplinary use and the application of the concept of the relationship outside of the psychotherapy/counselling dyad of professional therapist and patient. The alliance is not, in and of itself, represented as an intervention, rather it is the vehicle within which therapeutic gain may be facilitated [13, 17]. In this respect the model offers a synergistic view of the technical and process elements of therapeutic interactions (and case management); the two are reconciled within the context of the alliance rather than being perceived as discrete and unrelated entities.
Bordin describes the three major elements of the alliance as (i) tasks, the collaborative endorsement of the intratherapy activities (includes an understanding of what is required of each of the parties in the performance of these tasks, and an appreciation of the relevancy of the tasks to the therapy process); (ii) goals, the mutual agreement and valuing of the outcomes of the therapy; and (iii) bonds, that encompass the complex elements of attachment between the patient and clinician such as trust, empathy, personal liking and valuing. These elements reflect both the generic problem-solving nature of case management and the relationship element central to the more clinically or assertive based models. These components of the alliance are inherently flexible, with differing emphasis on each being possible in the various phases of therapy and within the differing theoretical orientation of the clinician [13, 15]. Bordin's model, therefore, has the potential to integrate more easily into both the multidisciplinary nature and the processes of case management. Additionally, the model has a clearly articulated theoretical foundation with a purposedesigned instrument set capable of measuring the constructs [13–15, 21]; an important consideration from a research perspective. The term ‘alliance’ is used from this point when referring to the therapeutic relationship.
Measurement
Studies of alliance to date have drawn on a basic set or ‘family’ of measurement tools related to the different theories [19, 20, 21, 22]. Various formats which measure the alliance from clinician, patient and observer perspectives are available, although not all sets include all formats. The instruments share two core concepts of the alliance, regardless of the theoretical orientation, the collaborative aspect of patient–clinician in the treatment process (although this may be represented differently) and the personal attachment or bond between the two [17]. Discrete aspects such as patient/clinician contribution to the alliance, patient participation in therapy and agreement on goals/tasks are also variously measured. There was no consistency among the scales as to the various weightings or emphasis given to these differing aspects [21] though high correlation between the scales has been demonstrated [21, 23, 24] indicating that they are measuring a common construct (the alliance) in addition to discrete aspects. The general development of these scales has been reviewed elsewhere [19, 21, 22, 25]; an outline only of the instrument set that measures the Bordin Working Alliance follows.
The Working Alliance Inventory (WAI) was developed from Bordin's model [15, 21] and aimed to capture the pantheoretical or generic factor(s) non-specific to therapy or technique, making it applicable to all therapies.
The instrument set comprises a patient (WAI-C) therapist (WAI-T) and observer form (WAI-O) and measures the tasks, goals and bond components of the alliance. It yields a global rating of the alliance in addition to subscores for the three components. Each item was rated on a 7-point scale with 1 = never to 7 = always. Reliability and validity have been established [15, 21, 24, 26].
Method
A computer search of English language articles using electronic databases MEDLINE, PsychINFO and Social Sciences Index was conducted for the period 1986–2000 using the search words ‘therapeutic relationship’, ‘therapeutic alliance’, ‘working alliance’, combining with ‘mental illness’, ‘psychiatry’, ‘community mental health’ and ‘case management’. A total of 84 articles and two texts were located. Abstracts were reviewed as available; otherwise titles were used to select articles relevant, but not limited, to psychiatry and community case management. Bibliographies from located papers were reviewed for relevant material; this extended the search period from 1979 onwards. An updated search was conducted in 2001–2002 using the terms ‘therapeutic relationship/alliance’ and ‘psychiatry and community mental health’, limiting to updates; to date this has not yielded new publications within this field. This literature review focuses on the published articles within mental health/psychiatry and case management, which used a validated measure of the alliance and outcome(s) measure (Table 2 and 3). Outcome measures included patient outcomes such as symptomatology, global functioning, quality-of-life, community living skill and medication compliance. Service outcomes such as hospitalization and programme attrition were also included (retention pertaining to the treatment period as distinct from continued case management). One article was excluded [27] as no psychometric data was reported for the 5-item visual analogue scale used as the alliance measure.
Methodology and summary: meta-anlaytic studies
Methodology and summary: psychiatry and case management studies (ranked for methodological rigour)
Results
Literature reviewed
Two notable meta-analyses [19, 20] a decade apart and spanning 20 years of research on the alliance were located. They analysed a total of 103 studies that met specific criteria and were primarily psychotherapeutic in orientation. All studies included outcome measures. A summary is provided in Table 2.
Four articles from general psychiatry [28–31] and four from case management [8, 32–34] were also found. Two of the case management studies [32, 33] and one general psychiatry study [30] were excluded from this review, as they did not include patient health outcome measures. Methodology details and a summary of the general psychiatry and case management studies are provided in Table 3.
Relevance of alliance to case management outcome
The primary evidence for the effect of the alliance on outcomes arises from the general psychotherapy literature; as few studies have been done in the case management field, relevant evidence from the psychotherapy literature was reviewed first.
Meta-analytic studies
Table 2 provides a summary of the methodology and pertinent findings of these two studies. The methodology used in both studies was equivalent; the second study [19] used the same inclusion criteria as the first [20] and also included the studies from the original meta-analysis. Two studies [28, 34] from the current review were included in the metaanalyses.
The two meta-analyses [19, 20] found a mean effect size (ES) ranging from 0.26 [20] to 0.22 [19] indicating that change in outcome, as a function of measured therapeutic alliance was modest. Although the ES was not large, the authors state that it was comparative to values reported for other psychotherapy variables. Effect size erred on the conservative side, as all relational analyses were included; where no ES was reported or was non-significant, this was set at zero [19]. The relation of outcome to alliance was consistent and not reducible by moderator variables such as type of outcome measure, the rater, time measured, methodology or type of psychotherapy [19, 20]. The WAI was used in seven of the 24 studies in the first review and was the most frequently used (n = 22) in the second.
Non-case management studies
Table 3 summarizes the methodology, findings and comparative ranking based on the methodological strength of the studies reviewed.
Three studies in general psychiatry were located that met the selection criteria. The setting for these studies varied; one [31] was in an inpatient unit in a Swedish facility – no further details were provided (ranking 3). The second study [29] was a rehabilitation programme in a Canadian public psychiatric facility within a research institute (ranking 4). The third study [28] (also included in the Horvath meta-analysis [20]) was undertaken as part of the Boston Psychotherapy Project with patients recruited from hospital admission and followed in the community for 2 years; no details regards public/private status was provided (ranking 1). Duration and randomization varied across the studies, with schizophrenia being the primary diagnosis.
All the above studies [28, 29, 31] showed significant correlation between the alliance and selected outcomes, however, the Frank [28] study was the most comprehensive and of the most interest in terms of the relationship between alliance and outcome and is discussed below. The alliance measure, the Psychotherapy Status Report, was purpose-designed and reported psychometrics established reliability and validity.
Frank [28] examined a comprehensive range of outcomes (see Table 3) and used cluster analysis to identify an index of overall outcome from a profile of all the measures of functioning taken over the two-year period. The alliance measure used was that of ‘active engagement’ (AE). This was a composite score of the alliance measure taken over the study period. The AE was set at ≤ 2, 3 and ≥ 4, good, fair and poor, respectively (the lower the AE score the more actively engaged the patient) [28]. The alliance demonstrated predictive power for both outcome and continuance in therapy.
Outcome
The relationship between early alliance (measured at 6 months) and outcome at two years demonstrated the following: those with ‘good alliance’, 77.8% (n = 14) also had good outcome and 22.2% (n = 4) had poor outcome. Those with ‘fair alliance’, 72% (n = 16) also had good outcome and 27.3% (n = 6) had poor outcome. Those with ‘poor alliance’, 25% (n = 2) had good outcomes and 75% (n = 6) had poor outcome.
Continuance in therapy
At 1 month the majority, 65 (51.2%), of patients had formed only a ‘fair’ alliance, with 44 (34.6%) patients who rated a ‘poor’ alliance and only 18 (14.2%) patients who rated a ‘good’ alliance. This picture did not change significantly at 3 months when active psychosis had largely been resolved.
The 6 month point showed a significant increase in patients with ‘good’ alliances over baseline, however, they were still in the minority (29.8%). Most of the 37 (71.8%) patients with a ‘poor’ alliance rating at 6 months left treatment within the next 3 months, only 2 (5.1%) progressed to a ‘good’ relationship [28]. The predictive power of the alliance rating for continuance in therapy held after testing for attrition [28].
Medication Compliance
Alliance ratings were also significantly correlated with increased medication compliance, with 73.9% (17) of patients with a ‘good’ alliance fully compliant with medication over 1.5 years of the study. In contrast, 74.2% (23) with ‘fair’ alliance and 72.2% with ‘poor’ alliance were non-compliant for the same period. Frank [28] noted that this represented a significant increase in non-compliance for those patients with ‘fair’ or ‘poor’ alliance over baseline compliance rates. For those with a ‘good’ alliance, compliance with medication held regardless of other variables such as type of medication, level of symptomatology and locus of care (hospital or community). Furthermore, the therapeutic gains by patients with positive alliances were achieved with significantly reduced neuroleptic medication dosage. Patients also spent less time in hospital although admissions were not significantly less frequent [28].
Pretreatment variables were not specifically examined, however, as detailed above, the patients, all with a diagnosis of schizophrenia, took longer to form alliances than did those in conventional psychotherapy. Patient alliance rating was not reported.
The Gehrs and Goering study [29] found significant correlation between patient and therapist alliance scores at both data collection points, between therapist Modified Goal Attainment Scale (M-GAS) scores at times 1 and 2, and patient alliance and patient M-GAS scores at time 2. No significant correlation was found for patient/therapist scores on the Problem List at time 1, but was demonstrated for therapist scores at time 2. There were no significant findings for changes between time 1 and 2 [29].
The Svensson study [31] did not examine the relationship between patient and therapist alliance ratings. Significant correlation was found between the therapist scores of the alliance and residual change scores for therapist-rated General Assessment of Functioning. None were found for patient alliance ratings and outcome.
Pretreatment variables were examined for effect on initial alliance and included social contacts, psychiatric symptoms, employment and inpatient service use. For patients, significant correlation was found between their initial alliance rating and level of target complaints. Therapist ratings of initial alliance were significantly correlated with patients' preadmission working ability. A subgrouping of patients based on bed days for 18 months prior to study entry showed that those with an excess of 165 days had significantly poorer initial alliance. No other significant correlations between initial alliance and patient characteristics were found.
Case management studies
Both of these studies were done in a community-based setting in the US. The Neale and Rosenheck [8] study was a substudy of a multisite project of veterans receiving intensive community case management (ranking 2). The Solomon and Draine [34] study was also part of a larger randomized-controlled study of patients receiving intensive case management in a community setting (ranking 2). Study participants met local requirements for intensive case management. In the Neale and Rosenheck study [8] patients were classed as having ‘serious mental illness’(not defined), however, 71% (n = 102) had schizophrenia, psychosis or major affective disorder and high hospitalization (not defined). In the Solomon et al. [34] study patients were classed a having a ‘major mental illness’ (not defined), diagnostic categories or percentage of patients with particular diagnosis not stated, and high hospitalization (defined as 60 days or more in the previous 2 years).
The measure of alliance was taken only once at the 2-year point in the above studies [8, 34]. Both found significant correlation between the alliance and outcomes at 2 years using regression analysis.
Neale and Rosenheck [8] examined pretreatment variables which included the number of inpatient days 12 months prior to study entry, measures of social competence (an objective composite measure derived from age, marital status, employment history and educational level) and global functioning. No effect for these on case manager or patient perceived outcome relative to improvement, benefit or satisfaction with services was found. Social competence accounted for some of the variation in correlations of increased symptom severity at 2 years, although it was case manager alliance that accounted for most of this variance.
This study (n = 143) found patient alliance ratings (mean = 197, SD = 29) significantly higher than those of the case manager (CM) ratings (mean = 187, SD = 29), although there was moderate correlation between the two. Case manager alliance was also strongly correlated with a greater number of outcomes, such as global functioning, symptom severity, community living skills and perceived outcome, than was patient alliance rating, which was only correlated with positive patient-perceived outcome. Hospitalization during the study period was not correlated with CM or patient alliance ratings and was identified as being primarily the result of preprogramme hospitalization patterns.
The Solomon and Draine study [34] (n = 90) showed that the alliance scores were high at the 2-year point for both patient and CM. The possible range was 36–252 with patients' average scores at 202.9 (SD = 35.1) and case managers at 200.9 (SD = 28.4). There was moderate correlation between the patient and case manager total and subscale scores. Both scores significantly predicted positive outcomes in quality of life, symptomatology, attitudes to medication compliance and satisfaction with services. The stronger the patient perceived the alliance with their case manager, the more positive was their attitude toward medication compliance.
Discussion
Overview
Overall, the psychotherapy literature to date supports the conclusion that the alliance was a ‘robust’ predictor of outcome [20]. A moderate, but consistent effect for alliance on outcome was found regardless of moderator variables such as type of therapy, time, interventions used, rater or type of outcome measure used and supported the link between positive patient–therapist alliance and improved outcome [19, 20]. The case management literature was small and studies were retrospective, however, they also demonstrated a significant correlation between the strength of the alliance and improved outcomes such as reduced symptom severity, improved global functioning and higher community living skills, improved quality of life, medication compliance and patient satisfaction with treatment. Yet, it must be noted that in one study in particular [8] these outcomes were primarily related to case manager ratings of alliance rather than patients. Several authors [28, 29, 31] noted that there were relatively few studies involving patients with psychotic illness and that there was little research into the process by which positive results are achieved in psychiatric rehabilitation interventions [29].
Methodology issues
Some methodological bias was evident in two studies; sample [29] and instrumentation bias [31]. Patients were either ‘actively engaged’ in therapy, potentially with an existing alliance at time of entry to the study [29] or assessed for motivation and collaboration with treatment prior to entry [31]. Additionally, different alliance measures were used for patients and therapists; no intercorrelational data was provided for the measures [31].
Generalizability was also of concern in the above studies. The sample population of Frank et al. [28] consisted of ‘non-chronic’ patients' aged between 18 and 35 with a diagnosis of schizophrenia who had minimal prior treatment. However, they were required to have functioned in a major role, outside of the hospital, without medication, for at least 4 months out of the previous 2 years; this is not typical of patients in case management or community mental health services in Australia. The settings and programmes for the two non-case management studies [29, 31] were specialized and, although the two case management studies were more amenable to generalization, they too have some constraints. Both had a patient population comprised of veterans [8, 34], with one study involving both consumer and non-consumer case management teams [34], again not mainstream community psychiatry. However, these factors aside, the findings for effects of alliance on outcome are broadly consistent with other findings, therefore may be used with confidence as basis for further study in community psychiatry and case management.
Meta-analytic studies
While the methodology used in both studies was the same there was some difference in the statistical procedures used to estimate the weighted effect size in the second study. We interpreted this as a feature of advances in techniques between the two time periods of the analyses, rather than any marked difference in end result.
An interesting point was that the first meta-analysis tested for homogeneity of the ES in the 20 studies, and concluded that there was more variability than could have been accounted for by chance alone. A suggested reason was that it might be related to the effect of moderator variables, such as patient factors, treatment differences, instruments or outcome variables. However, as no significant differences for any of these factors was found, this would seem unlikely.
Horvath [20] also found greater homogeneity among ‘families’ of instruments measuring the alliance and more heterogeneity across instrument families. He suggests that this reflects the ‘unique’ aspects measured by some instruments rather than disparity in the concept of the alliance and notes that previous studies have found that the various instruments are highly correlated (see the measurement section of this article). In contrast, the second meta-analysis [19] found homogeneity for ES and therefore concluded that no moderator variables were operating, so did not test for instrumentation effects. However, this difference in finding for effect size and homogeneity/heterogeneity remains an unexplained anomaly between the two studies, all the more puzzling given the overlap of studies in the meta-analyses.
Key issues
Issues pertinent to the alliance and case management arising from the literature reviewed are discussed below, included are pretreatment variables, patient characteristics, congruence between patient and case manager ratings of alliance, case management models, medication compliance and applicability to the Australian community mental health setting.
Pretreatment variables
Diagnosis as a pretreatment variable was not directly examined in any of the studies in this review. However, one non-case management study found that comparatively, patients with a psychotic illness had greater difficulty in forming relationships than patients in general psychotherapy, and that an extended period was needed to achieve even a fair relationship [28]. This study also found that the first 6 months were critical for alliance development [28]. A good alliance at this time correlated with significantly greater compliance with medication, better programme retention, and better outcomes after 2 years, with reduced medication dosage. Patients not achieving this by 6 months were considered unlikely to improve over a longer period. Patients with poor alliance at this time were more likely to ‘drop out’ of therapy or have poorer outcomes and more hospitalization.
The time period for development of a positive alliance in a conventional psychotherapy patient population was in early stage therapy around the third to fifth session or approximately 1-month [28]. This finding has clear clinical significance for a population of patients with a psychotic illness being case managed in community settings. It demonstrates the importance of prompt access to a stable case manager; although a positive alliance may not have developed by 3 months from initial engagement, when patients may be actively psychotic, it is possible for a good relationship to develop from this time. However, if not developed by 6 months from engagement, case managers may need to apply targeted relational techniques to foster alliance development or a change in case managers may be considered.
A further point of interest regarding patients with a psychotic illness was the finding that patients with a diagnosis of schizophrenia rated a poorer alliance at discharge compared with other patients in the study [31]. There was no difference in therapist ratings of this group of patients and no explanation was suggested for this finding. Interestingly, mental health (also highly associated with the quality of interpersonal relationships) was one of the influencing factors identified as important in the formation of the alliance and its predictive capacity on outcome in general psychotherapy [21, 35].
Patient characteristics that may influence the alliance and outcome were examined in one case management study only [8]; two only were significant (social competence and hospitalization). Social competence, including working ability was also significant in one of the noncase management studies [31]. Although pretreatment factors were not reported in the remaining studies Frank [28] noted the importance of examining contextual factors including patient/therapist characteristics in future research.
The finding on social competence was interesting as factors of social relationships/competence were identified as important in the formation of the alliance in general psychotherapy [21, 35]. Furthermore, social competence and its potential impact on alliance formation and outcome was particularly notable in view of the previous discussion of the difficulty in building effective alliances in patients with a psychotic illness. Many patients in community case management have psychotic illnesses and lack social skills and networks that allow for the development and or experience of stable and productive relationships [29, 36, 37]. Deliberate relational techniques may be required to promote an effective alliance with this subgroup of patients [29], in addition to the valuing, by case managers, of the relationship as at least of equal importance as medical management in the overall therapeutic management of the patient.
Also of interest here was the finding for a group of case managed patients who did not complete the patient measure of the alliance (n = 23) [8]. These patients scored significantly lower on measures of global functioning and community living skills at entry to the study and in case manager rated alliance at 2 years. It would be interesting to know what this groups' score was on the pretreatment variables.
Congruence of patient and case manager ratings of alliance and outcome
Alliance ratings
Both case management studies [8, 34] found moderate but significant correlation between patient and case manager scores of the alliance (both used WAI – patient and therapist versions). One non-case management study [29] examined this variable and found moderately strong correlation.
Alliance and outcome
Neale [8] reported a somewhat anomalous finding – case manager rated alliance was associated with more outcomes than was the patient measures (6 measures for CM vs 1 for patients). The authors noted that this was greater than the usual findings in general psychotherapy research (where patient ratings are more predictive of outcome than therapists). They suggested that this increased finding for case manager alliance and outcome may be the result of the patient/case manager dyad in the in vivo environment and the intensity of contact possible within the intensive case management model in this study. However, this seems somewhat contradictory, as one would expect that this contextual factor would lead to greater congruence between patient and case manager ratings of outcome.
One non-case management study [29] also found therapist rated alliance associated with more outcomes than was the patient rating. The authors of that study suggest that this may be related to perceptual differences of patients and therapist both of the alliance itself and the rate of clinical progress, or the result of rater bias by therapists. Another possible explanation given related to the conceptualization of Bordin's model (the unidirectional nature of the relationship) which may be differentially influenced by perceptions of the alliance by patients and therapist. The authors suggested that this might vary between different patient/therapist dyads and by differing measures of what constitutes an effective alliance. Consequently, they suggested that Bordin's model might need ‘re-visiting’ in the light of this finding. However, we argue that this finding does not negate the unidirectional nature of the alliance, but rather represents another principle of the model, that of ‘goodness of fit’. This principle represents the strength of the alliance as a function of the ‘goodness of fit’ between the patient and therapist personal characteristics and the unique demands of that particular alliance; all of which is influenced by the differing psychotherapeutic approaches [13]. More evidence would be needed to warrant challenges to the conceptualization of the model than was provided by this one study.
The above study [29] also raised the contentious issue of insight and judgement in persons with a psychotic illness when responding to self-report questionnaires. The study patients all had a chronic psychotic illness, however, high congruence was found between the therapist and patient alliance ratings and the outcome rating for the Modified Goal Attainment Scale (M-GAS), with the authors stating that this finding indicated that ‘perceptual differences’ were not problematic in this study. This would seem somewhat contradictory given the argument advanced above by those authors relative to the greater number of outcomes associated with therapist ratings. Caution is required in applying this finding generally to the field of mental health, as there have been too few studies to draw any conclusive findings regarding effect of psychotic illness on insight and capacity for judgement relative to self-report.
The general finding in the current review, including those of the meta-analytical studies [19, 20] was that therapist scales across all instruments have significantly poorer predictive power than do patients or observer ratings [20, 35]. Horvath [20] suggested that this might be due to the patients' perceptions being based on both past and current observations and their ability to make a judgement of the current relationship within the context of their previous levels of collaboration or engagement. One explanation suggested for the therapist findings being less predictive was that the therapist may ‘overestimate’ the quality of the relationship in the early phase, either through misinterpreting over-compliant behaviour as collaboration or through countertransference [20]. Various other explanations relative to patient or clinician perceptual differences or methodological issues together with issues of countertransference have also been suggested by others to account for this difference [15, 17, 20, 24].
Models of case management
Closely related to the above discussion, is the influence of contextual factors such as the model of case management, on the therapeutic milieu. As discussed above, Neale and Rosenheck [8] attributed the greater range of outcome associated with case manager alliance ratings as a potential function of the intensive case management model operating in that study. Solomon et al. [34] also noted that subjective outcomes were best predicted by patient alliance, whereas, case management models generally tend to focus on objective outcomes, such as hospitalization, symptomatology and treatment compliance. He noted that there have been mixed results reported for achievements in these objective areas and that patients may be differentially focused on other aspects of their lives, particularly the more subjective aspects as indicated by the study findings. This comment was not only relative to case management, but also to the non-case management studies in this review as most have focused on objective measures of outcome (Table 3). Further weight was added to this argument in the case management study by Neale and Rosenheck [8] in which objective pretreatment variables had no significant effect on subjective perceptions of benefit, improvement or satisfaction with treatment for patients or case managers. Consequently, Solomon et al. [34] suggest that the alliance may be a particularly useful vehicle for improving the patient's subjective perceptions of community living. However, they acknowledged that subjective outcomes were more of a focus in clinical case management models, the use of which may need more professionally qualified staff than generalist models.
Medication compliance
The issue of medication compliance is a long-standing one of major concern to service providers and practitioners in the mental health field and one closely linked to recidivism in the patient with psychotic illness [38–44]. The Solomon [34] case management study demonstrated a significant relationship between medication compliance and the patient rated alliance, although they noted that this was an attitudinal measure rather than a direct compliance measure. Only one non-case management study examined this variable [28] and found a significant association with alliance and compliance and reduced medication dosage. This was a very pertinent finding as case managers spend a good proportion of their time monitoring medication, and some services devote after-hours teams to the delivery and supervision of medication to patients. A better understanding of the origins and growth of the Working Alliance may aid in the effective delivery of these important outcomes.
Relevance of findings to the Australian community health setting
Issues of relevance of findings to the Australian setting arise from three sources in this review. First, the diagnostic profile of the patients in the studies reviewed (patients with a psychotic illness, primarily schizophrenia). Second, the models of case management reported (intensive-based models) and finally, the potential difference in the nature of the help-seeking behaviour between patients in psychiatric services as opposed to general psychotherapy.
Case management services
Intensive case management is not the ‘norm’ by which the majority of services are provided to patients. There was no published data on the models of case management currently used in the Australian community mental health sector, however, anecdotal evidence would suggest that most community services in Australia are likely to be characterized by an eclectic mix. Some services may provide a more clinical or intensive component via small, specialized units that may also include outreach services through mobile treatment teams, crisis teams and after-hours services. The majority of case management however, would seem to be delivered by a multidisciplinary team of case mangers with individual patient loads of 20–30, who attempt to provide a gamut of services depending on (often conflicting) patient needs.
Psychotic illness
Again, national data on the percentage of patients with a psychotic illness being treated in community case management was not currently available. However, some recent Australian publications provide pertinent data [45–47], key points of which are outlined in Table 4.
Key points relative to persons with a psychotic illness
As the above data would indicate, patients with a psychotic illness might comprise a small percentage of the mentally ill population, however, they represent a resource intensive group, and are considered a major public health issue in Australia [48]. The majority of patients with a psychotic illness were living in the community and required long-term support and intervention; many needs of this group were unmet under current community-based services [47, 48].
A recently released report by Access Economics, commissioned by SANE Australia [47] makes sobering reading and highlights the direct and indirect costs of persons with schizophrenia to the individual, families and society. The human cost of schizophrenia in 2001 was 22.6 healthy life-years, including 129 lives lost as a consequence of suicide. The total financial cost was $1.85 billion.
Helpseeking
As discussed earlier, patients in general psychotherapy are usually voluntarily seeking help and enter a relationship motivated to engage; this is not necessarily the case with patients in mental health, who may be under a court order for treatment. However, it was not possible to objectively determine the precise relevance of this factor to the application of the concept to the Australian service setting as no national data on legal status for communitybased services was available. Furthermore, State and Territory differences exist in what comprises an ‘involuntary order’ [46]. However, the State of Victoria publishes annual statistics for the Mental Health Review Board [49], and this gives some insight into the area (see Table 4.
As this data indicates, the majority of persons presenting to the court had a psychotic illness and were being managed in the community. Additionally, patients who are admitted to hospital under an involuntary status may also be discharged to community care and case management with the involuntary detention order still in effect and would represent additional numbers to those indicated above. Thus, the legal status of the patient may represent an additional ‘barrier’ to engagement if the patient is resistive to the mandatory nature of their treatment. Patients with involuntary treatment orders therefore represent an additional challenge to the case manager attempting to forge a therapeutic alliance with the patient. No information was provided on this factor in any of the studies reviewed, however, it is of clinical importance to future research in the Australian setting.
The above findings on the current state of patients with a psychotic illness being managed in the community make the findings of the current review on the benefits of the Working Alliance on patient outcome for this group very relevant to the Australian setting. Mechanisms that facilitate better engagement with service providers and lead to improved outcomes has potential benefits for patients, service providers and society at large.
Clinical and research implications
Clinical
The Working Alliance is integral to both service delivery systems and clinical practice. It provides a focus on patient outcomes as opposed to systemic outcomes, as it is a collaborative process that centres on patient needs and goals, versus clinician generated goals.
A practical example of this is the potential the alliance has to improve medication compliance. This is a major issue for the patient and for service delivery as low compliance equates with poorer outcomes and increased symptomatology, leading to increased service use including hospitalization, and increased costs. The Working Alliance offers avenues away from coercion and power factors in medication compliance. Emphasis can be put on practical patient outcomes (what patients want from treatment and how medication can help them achieve this). Medication becomes a vehicle for change versus externally imposed intervention. In the process the individual is potentially empowered, with gains in problem solving skills and self-esteem. The alliance process is one that promotes partnerships with patients and facilitates selfmanagement through active engagement of the patient in the treatment process. Many patients with serious mental illness have been unintentionally disempowered by the very systems and services provided to support them; the Working Alliance offers an alternative. This will not, however, be achieved without organizational investment and commitment to its human resources and to the review of policies and practices, particularly issues such as manageable workloads and continuity of care, which promote this approach to patient-focused service delivery.
Research
Of particular importance to alliance research in case management was the possible maturational effect inherent in the retrospective studies of the alliance. In measuring the alliance after 2 years of treatment, both clinical improvement and a strong alliance could be expected [34]. The correlational findings do not infer causality, and treatment gains over the 2 years may have influenced the alliance rather than the converse [28]. Prospective designs with periodic assessment of the alliance to track development within a more inclusive framework were recommended [34].
What does this mean for Australia?
In light of the findings of this review, research into the Working Alliance in Australian community mental health is warranted in order to provide data on the role and function of the alliance in achieving improved patient outcomes. As evidenced by the current review, prospective studies that monitor the development of the alliance within an integrated model of case manager process and service delivery are desirable.
What needs to be included?
Factors such as patient and case manager characteristics, and the key patient and service outcomes identified from this review should be included. Other potential factors which may impact on the alliance, such as the skill/competencies of clinicians, styles of engagement of patient and clinician, personality factors and interventions that promote alliance development, may help to identify the therapeutic elements inherent in case management that have proved elusive in research to date.
Special considerations
Any research that includes patients with a psychotic illness would need to consider the restraints imposed by the accompanying symptomatology, and the longer developmental phase which may be required in forming productive alliances with this group of patients. These would influence study design, particularly in relation to the length of study and data collection points and recruitment processes and methods. Outcome measurement would be similarly influenced as any changes, particularly in subjective outcome, could be expected to be a function of time, with no significant gains obvious in the short term. Longitudinal studies would be the most optimal to meet these requirements. Given that the alliance is a dyadic one between patients and case managers, careful consideration to this point is needed in study design, particularly in the recruitment phase as active support of patients and case managers is required. Flexibility is also needed, with consideration given to case managers concerns regarding the potential impact of a third party impinging on the difficult initial engagement phase when many of the patients may be in an acute phase. Similarly, this is not an optimal time for recruiting study patients. These factors would need to be balanced against the desirability of early baseline data collection to allow for tracking of alliance development and providing optimal conditions for analysing effect on outcome.
What next?
Currently, a prospective longitudinal study into the Working Alliance between patients with serious and persistent mental illness and their case managers is in progress in an Australian community mental health service.
Summary
Although there was minimal evidence of the assumed link between the case manager relationship and improved outcome for patients with serious and persistent mental illness, this review clearly indicates that the alliance was definable, measurable and relevant to the core business of case management – improving patient outcome. Furthermore, the small pool of research in the field of case management showed encouraging results. Bordin's concept of the alliance shows potential as a pantheoretical variable across multidisciplinary settings, including that of case management, with the findings indicating its potential as a predictor of therapeutic outcome. Studies to date have primarily focused on patients with a psychotic illness who have been intensively case managed. Although this focus has issues of relevance to the current Australian setting, care would be needed in designing relevant research. However, patients with a psychotic illness are very clearly of particular concern in any community-based setting. They represent a comparatively small, yet highly significant proportion of persons with mental illness; current Australian research has demonstrated the high burden and cost associated with the management of such patients in the community. Potentially, the alliance has positive clinical and service benefits that ultimately can improve patient outcomes and service effectiveness. Research into the role and function of the alliance in the Australian community mental health field is essential. A prospective research project arising from the current review is in progress in a community health service as a foundation for future research to assist in identifying the therapeutic elements of the case manager process and case management services.