Practice-Friendly Research Review: Collaboration in Routine Care

Abstract

This article examines the research on patient-psychotherapist collaboration in ways that can inform and improve clinical practice. Clinical wisdom suggests and research supports the importance of goal consensus and collaboration; empirical support for this assertion is summarized and the implications for practice are provided. Then, we present a method of heightening collaboration through the use of assessment and feedback. Systematically monitoring psychological functioning, client perceptions of the therapeutic relationship, motivation and expectations of therapy, social support network, and untoward life events can enhance collaboration and ultimately treatment outcomes.

Goal consensus between therapist and client is considered an essential element of successful psychotherapy. Collaboration between therapist and client, a related concept, is also considered essential, as the process of reaching goals requires the therapist and client to agree on the necessary steps to reach such goals. These two elements are fundamental to treatment, making up two of the three elements of the working alliance (with the third element being a positive affective bond between the participants). Empirical support for these therapy elements has certainly been found, but is complicated by the fact that goals can readily change during treatment and require modification in the tasks over time.

To some degree, neither client nor therapist can predict at the inception of treatment if goals will change and in what ways, and if working together in and out of sessions will need to be modified. The natural flow of psychotherapy can therefore leave the participants occasionally at odds about what needs to happen now. To be sure, the fluid nature of therapy presents challenges for its empirical examination. In such research it is difficult to not only define the constructs of interest but also deal with the dynamics of a change process that include considerable variation over time.

In this article, we review the empirical evidence for the relationship between goal consensus and collaboration and treatment outcome. This summary is followed by suggestions for enhancing goal consensus and collaboration.

Research Review

A wide variety of overlapping reviews of psychotherapy relationship variables (such as the therapeutic alliance, group cohesion, empathy, positive regard) suggest that such elements of treatment are at least as important in contributing to patient change as specific treatments (Norcross & Wampold, 2010). Among such variables are goal consensus and collaboration. The most comprehensive summary of the empirical basis for these two related variables has been published by Tryon and Winograd (2011), whose recent meta-analysis considered 15 studies that examined the relation between goal consensus and outcome along with 19 studies that reported on collaboration and outcome. Although this rather small set of studies that met criteria for inclusion in their quantitative review is disappointing, it represented only findings from 2000 through 2009.

The 15-goal consensus studies produced a mean correlation of .34 with a 95% confidence interval of .23 to .45. To remind the reader, a mean correlation of .34 translates to an effect size (d) of .72 and indicates that therapy marked by high goal consensus would have 67% of the patients with a successful outcome, while therapy marked by low goal consensus would have about a 33% success rate. Similarly, these researchers found a correlation of .33 (confidence interval .25-.42) between collaboration and outcome across 19 studies. Clearly these data indicate that patients can improve despite relatively low levels of goal consensus and collaboration, but that improvement is much more probable if therapists and patients can work together to identify goals and steps to reach these goals. Most of this research is correlational in nature rather than experimental.

Typical studies report the correlation between a measure of collaboration and eventual change. Examples of research questions are as follows: Did clients turn in the homework that they agreed to? Does the percent of homework turned in correlate with a measure of mental health functioning/symptomatic improvement? For example, Simpson and colleagues (2011) examined the effects of adherence to between-session homework involving exposure exercises for two groups of patients being treated for obsessive-compulsive disorder (OCD). Outcome was judge-rated OCD intensity posttreatment correlated with therapist weekly ratings of homework completion. They found that outcome was correlated with the degree to which patients completed weekly homework assignments using linear regression that included other predictors of outcome. In such research, therapists can rate the quality of the homework, though this is not usually done.

It is important to keep in mind that the substantial correlations with outcome, found across the reported studies, are not indicative of a causal relationship. In fact, a limitation of this area of research is the failure of researchers to design studies that, at a minimum, could show that collaboration (such as homework assignments completed) preceded improvement, rather than the improvement preceding collaboration or leading to goal consensus. Empirical studies on goal consensus have not generally investigated the extent to which both outcome and goal consensus were caused by a third unstudied variable, like good interpersonal skills or the sense of coherence experienced by the patient that existed prior to the commencement of psychotherapy.

Some studies have attempted to investigate the degree to which homework compliance predates improvement. Burns and Spangler (2000) were the first to investigate the possible causal relationships through path analysis. They demonstrated that homework compliance preceded improvement rather than followed improvement in a sample of depressed patients.

Another challenge facing the scientific investigation of collaboration and goal consensus is one of operational definitions. In the area of goal consensus, it is not always easy to specify goals and agreement between therapist and client. Few studies of goal consensus have actually examined specific goals but instead rely on client perception of goal agreement. In an exception to this trend, Busseri and Tyler (2004) examined the relationship among actual client-therapist agreement on specific target problems and perceived agreement on goals based on working alliance ratings and their correlation with outcome. Therapist and client dyads completed target complaints lists and symptom ratings at three points over the course of therapy, while working alliance ratings were collected at the fourth and final sessions. Both target complaints and alliance at the fourth session of treatment contributed to the prediction of outcome, but agreement on targets was unrelated to client perception of goal agreement. These results suggest that it is important to examine actual goal agreement rather than relying only on general perception of goal agreement, as measured by the working alliance inventory.

Collaboration research over the last decade has been dominated by studies of behavioral and cognitive behavior therapies and homework compliance rather than more broadly examining collaboration across therapies. Nevertheless, a concentration of studies on the effects of homework on treatment outcome has accumulated in cognitive behavioral therapy (CBT). An important contribution to this area is the series of meta-analysis conducted by Kazantzis and colleagues (Kazantzis, Deane, & Ronan, 2000; Kazantzis, Whittington, & Dattilio, 2010). These authors examined a wide variety of studies (N = 45) using various research designs. Nine of these studies examined the effects of homework in experimental investigations that contrasted CBT treatments with and without homework. Although the study results were heterogeneous (g range -.19 to 1.11), the average effect (d) was .48, suggesting that 62% of treated patients who were given homework would have success compared with 38% successful cases where formal homework assignments were not made within the context of CBT. Homework makes an important contribution to CBT outcome, and this element of treatment is probably the most investigated aspect of the collaborative relationship.

The research discussed above highlights the significance of goal consensus and collaboration for positive patient outcomes. This evidence, although mostly correlational, suggests the importance of emphasizing therapists’ abilities to negotiate goals and methods for reaching these goals during clinical training and in supervision, regardless of the type of psychotherapy that is offered to clients. This body of research has not shed much light on ways of maximizing consensus and collaboration in the context of the individual case. It is difficult to know, for instance, how to improve collaboration when it is insufficient.

One approach to increasing such collaboration may be derived through assessment of individual patients over the course of psychotherapy. In the following section, we summarize methods that may be used to assess client functioning, goal consensus, collaboration, and related variables to provide feedback to therapists for the purpose of ensuring greater collaboration and enhancing treatment outcomes. These methods can be applied to all clients, but seem particularly consequential for patients who are not deriving a benefit from therapy.

New Approaches to Increasing Collaboration

Using outcome assessment as a tool to more fully understand the patient's struggles as well as his or her progress in therapy fits with the goals of evidence-based practice (APA, 2005). Specifically, to maintain a high quality of patient care, practitioners are advised to consider, on a case-by-case basis, whether a particular treatment is working for an individual patient. This question can be addressed through the use of systematic measurement of mental health that evaluates patient progress over the course of their treatment.

The material presented here is atheoretical and intended to be applied regardless of theoretical orientation. It focuses on a universal problem across psychotherapies by targeting patients at risk for treatment failure. The methods discussed bring therapist and patient together to address the problem of poor progress in therapy. Finally, it relies on formal assessment to enhance collaborative work.

Problems in Psychotherapy

Decades of research have provided evidence of both the substantial benefits of psychotherapy and the need for improvements. In a summary of clinical trials research, about two thirds of adults who entered treatment either reliably improved or recovered after 14 sessions of treatment, leaving about one third of the patients unimproved or worse off (Hansen, Lambert, & Forman, 2002). In routine practice settings, clients did not fare as well as those in clinical trials. Only about one third of these clients showed reliable improvement or recovery. The largest percentage of clients in these studies showed no change, and 6%–10% were considered treatment failures. In children and adolescents the situation is probably even more dire, with deterioration rates closer to 15%–23% (Warren, Nelson, & Burlingame, 2009). These data suggest the importance of tracking individual client progress, the implicit if not explicit goal of clients and therapists, during the course of treatment to identify patients who are either failing to respond to treatment or deteriorating in treatment and to do this before termination of services occurs.

Identifying At-Risk Cases

It is important that this monitoring be done in a systematic and standardized way rather than relying on clinical judgment. Just as physicians monitor blood pressure or blood sugar levels to manage cardiovascular problems or diabetes, mental health status can also be monitored and treatment modified based on the responsiveness of clients to treatment. Mental health vital signs can be used to identify the cases at risk for negative outcomes and to help clinicians and clients adjust their activities. Such clinical management rests on the assumption that therapists cannot identify poorly responding clients in a timely fashion and thereby take corrective actions.

The clearest example of clinician failures in this regard was provided by investigators (Hannan et al., 2005) who examined therapist predictive accuracy. They asked 40 therapists at the end of each session with each of their clients if they believed the client would leave treatment in a deteriorated state and if the client was worse off at this particular session than when he or she entered treatment. The researchers expected that experienced clinicians would be more accurate in their judgments than trainees. Therapists were aware of the purpose of the study, understanding it to be a contest between experienced and less experienced professionals compared with statistical methods that they had used in the recent past. They also understood that there was no consequence to the client for making any prediction, as the research was aimed at understanding how well clinicians could forecast negative final treatment outcome. They were aware that the dependent measure used to categorize patient change was the Outcome Questionnaire-45 (OQ-45) and they understood the cutoff scores for judging deterioration, but they did not have access to the patients’ OQ-45 scores. Therapists were reminded that the base rate for deterioration was likely to be 8%, so the phenomenon they were to predict was relatively rare, perhaps 1 in 10 of their clients. Most therapists had experience receiving predictive information in prior studies using the OQ-45.

During a 3-week period, predictions were made for 550 clients who participated in therapy sessions. Treatment continued as usual, and clients’ progress was followed until they terminated treatment. Although 40 clients had deteriorated by the end of their treatment, only 3 out of 550 (.01%) were predicted by their therapist to leave treatment worse off than when they began. Of these three, only one was an accurate prediction, with the client actually leaving treatment deteriorated. Despite being informed of the base rate information for deterioration (which closely matched the actual deterioration rate of 7.3% for this sample), therapists showed an inability or perhaps unwillingness to accurately forecast negative outcome for their clients.

To become more aware of pending treatment failure, actuarial methods that take into account massive amounts of information about the treatment response of thousands of patients across thousands of therapists can be used to identify probable treatment failures. For example, in the Hannan et al. (2005) study, use of these actuarial predictive methods led to the identification of 36 of the 40 deteriorated cases (recall that the therapists identified only one of 40 such cases). In another study, of 492 clients in treatment at a university counseling center, 36 (7.3%) were reliably worse/deteriorated at termination. The actuarial method correctly identified all 36 (100%), most of whom (86%) were identified by the third treatment session (Lambert et al., 2002), Thus, the actuarial method can effectively identifying clients who went on to deteriorate and such a prediction could be made early in the treatment.

Monitoring Treatment and Providing Feedback

The most important aspect of tracking change on a session-by-session basis is assessing whether client scores tend to increase (get worse), stay the same, or decrease in relation to the intake score. Comparisons of patients’ outcome when session-by-session functioning has been provided to clinicians and when clinicians practice as usual have been studied in six large randomized controlled studies to evaluate the impact of using the OQ System (OQ-45) to assess and modify ongoing treatment response (Harmon et al., 2007; Hawkins, Lambert, Vermeersch, Slade, & Tuttle 2004; Lambert et al., 2001, 2002; Slade et al., 2008: Whipple et al., 2003). These studies each required about 1 year of data collection and included session-by-session measurement of over 5,000 patients. All six of the studies assessed the effectiveness of providing therapists with session-by-session progress data as measured by the OQ-45, with particular focus on identifying patients who were not responding well to treatment (signal-alarm cases).

Progress data were supplied in the form of a graph of OQ-45 scores detailing patient improvement and warning messages when improvement was not of the expected magnitude (progress was at the 15^th percentile or worse). Additionally, three of the studies assessed the effect of providing both therapists and patients with OQ-45 progress information, and three of the studies assessed the effect of providing therapists with additional feedback regarding the patient's assessment of the therapeutic relationship, motivation, degree of social support, and life events (termed clinical support tools [CST] feedback). These latter four assessments were provided in concert with OQ-45 progress information when it was deemed that the patient was not progressing in treatment as well as expected and, in fact, predicted to leave treatment deteriorated. Feedback using the CST was accompanied with suggestions for action that the therapist could consider and implement or discuss with the client.

These six studies included a number of commonalities: (a) patients were randomly assigned into control (no feedback) or experimental (feedback) groups at intake; (b) the same therapists who saw control condition patients also saw experimental condition patients, thus minimizing the possibility that measured differences are attributable to therapist effects; (c) the therapists represented a variety of treatment orientations, with the majority ascribing to cognitive behavioral and eclectic orientations; and (d) professional therapists represented about 50%–100% of the clinicians participating in each study, with the balance comprised of graduate student or postdoctoral trainees.

Figure 1 provides a sample progress feedback report. Note that the progress status is presented in the top right hand corner of the report and is “RED.” In the graphic display of progress, the patient's session-by-session OQ-45 score is provided in relation to a horizontal line, which indicates the cutoff for normal functioning 64/63 and a sloping line that indicates the patients expected progress based on 220 patients who began treatment at the same initial level of disturbance. The “RED” alert for this case indicates that only 10% or fewer patients have responded this poorly at this session of care. A message to the therapist is provided below the graph. In Figure 2 a progress graph and message is displayed and would be provided to the patient and then discussed with the patient.

Figure 1. Clinician feedback report.

Figure 2. Client feedback report.

Figure 3 presents the feedback graph based on the patient's responses to the Assessment for Signal Cases (an aspect of the CST). Note that feedback provides, in the top right-hand corner, subscale results which in this case indicates that the overall response to therapeutic alliance questions indicate an alliance that is at least a standard deviation below other patients ratings of their therapist. This is also true for the Motivation sub scale (also RED). On the left side of the feedback report specific item responses are highlighted because the answers are a standard deviation below the mean of patients receiving psychotherapy.

Figure 3. Feedback chart for the assessment of signal cases.

Figure 4. The effects of feedback-assisted psychotherapy for not-on-track patients compared with treatment as usual.

Note. NOT-NFb =treatment as usual for not-on-track cases; NOT-Fb = not-on-track cases whose therapists received progress feedback; NOT-Fb+CST = not-on-track cases whose therapists received progress feedback as well as clinical support tools; NOT-Th&PT = not-on-track cases in which therapists and patients received progress feedback; OT-Fb = on-track cases whose therapist received feedback; OT-NFb= on-track cases that received no feedback.

Feedback at the item level is deemed as important because it provides more actionable information than subscale feedback. The collaboration that emerges from this methodology invites clients who are off track to answer specific questions about their relationship with the therapist (as well as motivation, social support network, or disturbing life events), compare those perceptions with normative data, and make the assumption that there is a problem that needs solving when the answer is below expectations. For example in the case of George, Item 9 “My therapist seemed glad to see me” was rated “neutral” by the client and only about 15% of clients answer “neutral” or less. Thus, the therapist can contemplate what, if anything, to do about this fact. They can use the “Display Interventions Button” for suggestions to resolve problems with the therapeutic relationship.

These methods are seen as heightening the collaborative efforts by directing attention to the degree of improvement the client is experiencing, alerting both participants to unexpected negative change, analysis of the possible reasons for this negative change, and joint efforts to problem solve. We view the most important aspect of these feedback efforts as having therapists and clients face a lack of progress toward healthy functioning and addressing those factors that could be accounting for negative rather than positive change.

Effects on Client Outcomes

Table 1 provides a summary of outcomes from the six studies based on a meta-analytic review (Shimokawa, Lambert, & Smart, 2010). Each client's final treatment outcome was classified using Jacobson and Truax (2001) criteria for clinically significant change. As can be seen, the results of using these interventions are quite dramatic in regards to treatment-as-usual offered by the same clinicians.

Table 1. Clinical Significance Classification of Not-On-Track Patients by Treatment Condition

	Recovered or improved	No change	Deteriorated
Note NOT-NFb =Treatment as usual for not-on-track cases; NOT-Fb = not-on-track cases whose therapists received progress feedback; NOT-P/T = not-on-track cases in which therapists and patients received progress feedback; NOT-Fb+CST = not-on-track cases whose therapists received progress feedback as well as clinical support tools.
NOT-NFb
(n =318)	71(22%)	183(58%)	64(20%)
NOT-Fb
(n =263)	99(38%)	140(53%)	24(9%)
NOT-P/T-Fb
(n =177)	80(45%)	71(40%)	26(15%)
NOT-Fb+CST
(n =217)	114(53%)	91(42%)	12(6%)

The percent of patients who responded to treatment nearly doubled, and the deterioration rates dropped by two-thirds, among those who were predicted to deteriorate. Given the experimental design of the studies, this is deemed a causal effect of the feedback method.

For these methods to work, they facilitate a high degree of collaboration between client and therapist. First, the therapist is provided with a rationale for assessing client mental health functioning on a regular basis; for example, it is a vital sign lab test to track mental health. Factor analytic studies of the OQ-45 (Lambert et al., 2011) suggest that it dominantly measures subjective well-being or psychological distress and to a lesser degree interpersonal problems and performance of social roles. Just as in medicine, where lab test data can define normal functioning and allow physicians to see the consequences of interventions (and thereby manage disease), OQ-45 data allows clinicians to better manage psychological functioning. Second, clients must be willing to collaborate in the assessment and to provide accurate self-reports of life functioning. Third, therapists examine the lab test data and discuss them with patients. Finally, the participants make treatment and relationship adjustments as necessary. These methods help participants collaborate to maximize patient outcomes.

Further collaboration is facilitated through the use of the Clinical Support Tool that asks patients to offer feedback to clinicians about the agreement on goals, processes, the bond, expectations and motivation as well as problems in their extra therapy support network. Adjustments to therapy, such as bringing in family members or adding group treatment, require a flexible therapist and willing client working in collaboration. Without frank responses from clients and clinicians, who take time to reflect on the information provided through the Assessment of Signal Cases, there is no advantage to their use.

Other Systems

Several research groups have also developed systems for monitoring outcomes, although the evidence for improving outcomes lags behind that provided above. Barkham and colleagues (2010) created the Clinical Outcomes in Routine Evaluation (CORE) system. It is widely used in the United Kingdom to inform client care based on information gathered from mental health services.

The CORE comprises three independent tools. The CORE Outcome Measure (CORE-OM) is a 34-item, client self-report questionnaire administered before and after therapy (10 and 5 item versions are also commonly used for tracking). Ratings are rendered on a 5-point scale regarding how the person has been feeling over the last week. It provides a score indicating current global psychological distress. Progress is monitored and the information fed back to clinicians, if this is desired. Yet the strength of the system principally resides in the data it provides to administrators and managers of service delivery systems.

Kraus and Castonguay (2010) developed the TOPS system, which includes numerous evaluation tools covering child and adult functioning. The time needed for administration ranges from 2 to 25 minutes. TOPS, like CORE, has primarily focused on administrative uses, rather than feedback to therapists. Managers can examine progress throughout treatment and compare outcomes to appropriate benchmarks. The functioning of adults and children are quantified across a variety of areas and include diagnostic aids, historical information, and written statements of treatment goals. The report for clinicians includes ratings on 23 high-risk-related questions. Considerable emphasis is placed on the use of the report for treatment planning, the individualization of treatment goals, and tracking these goals. Client satisfaction, too, is measured and used as a quality assurance index.

TOPS requires users to send off forms for scoring and reporting. This procedure limits rapid turnaround of feedback for clinicians and the frequency with which the response to treatment can be tracked. The adult symptom scale is long, around 85 items, and has considerable redundancy within each area of disturbance (e.g., sleep, anxiety, mood). The length of TOPS does not make it ideal for tracking treatment response on a weekly or even biweekly basis, unless tracking is limited to specific subscales. For clients whose subscales are elevated, the authors do recommend using their tracking system each week. Overall, this practice has the advantage of targeting specific problems for specific clients, but it also carries the disadvantage of leaving untracked many items measuring symptoms. It is also hard to compare different treatments when different targets are being tracked.

In contrast to the preceding methods, Miller, Duncan, Sorrell, and Brown (2005) created an ultra-brief assessment package—the Partners for Change Outcome Management System (PCOMS). The PCOMS employs two, four-item (visual analogue) scales, one focusing on outcome (ORS) and the other aimed at assessing the therapeutic alliance (SRS). The measures are also available for use with children and adolescents (Duncan et al., 2006). Although brief, the ORS correlates modestly with other outcome measures such as the Symptom Checklist-90-R (SCL-90-R; .57), the CORE 34 (.67), and the OQ45 (.58; Duncan & Miller, 2008). It has the advantage of directly involving both clinician and client in the process of measuring and discussing both progress and the working relationship, thus ensuring a level of collaboration that is very high. During each session the therapist provides the measures to the client. Furthermore, as scoring takes place in the session, feedback is immediate. A commercially available web-based system (MyOutcomes.com, 2007) of administration, data collection, normative comparison, empirically based feedback messages, as well as aggregate statistics addressing a variety of effectiveness and efficiency variables are available to enhance the benefits of paper and pencil use of PCOMS.

The PCOMS system, using intake scores and progress at each session, provides information on anticipated treatment response. It also identifies clients whose improvement is falling short of expectations (Miller & Duncan, 2004). They have yet to examine accuracy of prediction of deterioration. Instead, they rely on sharing alliance and progress ratings with clients over the course of treatment. The goal is to ensure resolution of problems before they derail progress. Evidence is emerging, suggesting its positive effect on clients in both individual and couples therapy (Lambert & Shimokawa, 2011).

Conclusions

In general, we can expect all of these outcome-monitoring systems to enhance client-therapist collaboration and evaluation of goal achievement provided that they incorporate a warning system that identifies patients who are off track and deliver this information to therapists in a timely fashion (preferably instantaneously). In addition, problem-solving tools are highly likely to further bolster treatment effects.

Considerable empirical evidence supports the common sense belief that patient-therapist collaboration and agreement on treatment goals are predictive of and probably contributory to successful psychotherapy. The evidence is especially strong with regards to the power of homework completion to enhance outcomes in cognitive-behavior therapy.

Among the important new developments in psychotherapy is the use of brief assessments of client functioning to monitor patient progress and promote discussions of progress that falls short of expectations. Bringing clients and therapists together to face negative changes and their possible causes (e.g., goal agreement, task agreement, alliance ruptures) enhances collaboration and ultimate client outcome.