Search strategy

A comprehensive search of biomedical, psychological and social databases was conducted of English language papers published between 1998 and March 2009 from Australia, New Zealand, Canada, USA and Europe. These databases included MEDLINE, Embase, Psychinfo, Cinahl, ProQuest, Sociological Abstracts, Family and Society Plus, Meditext and all Evidence Based Medicine (EBM) Reviews. Inclusion criteria were based on the following operational definitions of primary mental health care and primary mental health care linkages.

Inclusion criteria

Studies which reported on factors that enable or hinder linkages were included. These used a range of descriptive study designs including surveys and questionnaires and qualitative studies which captured the experience of participants in the linkage. The qualitative studies chiefly used interviews, but also focus groups, observation of meetings and analysis of referral letters. Some of the papers were part of the randomised trials reported in the part one paper and provided valuable contextual information from these studies.

The restriction of a linkage to a clinical service meant that the review focused on factors at the clinical level. Hence, wider policy and program supports, such as national infrastructure were not included and this is a limitation of our review.

Data extraction & analysis

The data extraction template and coding framework are described in the part one companion paper. The papers coded as collaborative factors were further coded according to the enablers and barriers they described. Initial codes were agreed a priori based on the research team's knowledge and prior reading. The code list was further refined during data synthesis (JF & LH) to adjust for the use of different terms describing similar factors and to add new factors as required. Using this iterative process, the recurring themes on developmental factors in the studies were identified and refined through team and reference group discussions. The analysis comprised narrative synthesis whereby the factors were described and grouped by theme. This process involved tabulating the factors according to whether they were described as an enabler or a barrier to building service linkages (see additional file 1, appendix 1). We then theorised the inter-relationship between the factors through a graphical representation of the level at which they operate and whether they were a formation or a enabling activity (figure 1).

Since our review was designed to aid decision support for policy-making, the expertise of our policy maker and service director review group was particularly valuable as we considered recommendations for building service linkages. We first asked them to comment on how the findings from the international literature related to their experience in the Australian system and then, based on the factors identified, what they would recommend as acceptable, feasible and important actions to Australian health care policy makers.

Clinical level

At the clinical level we identified a set a factors describing activities that primary care and mental health clinicians would undertake together to form a partnership (linked service). We labelled these as "partnership formation activities" which we defined as the process of human organisation central to team building. We also identified further factors as support to partnership formation activities and we have mapped the inter-relationship of these factors in figure 1.

Three papers emphasised the need for all parties, mental health and primary care, to be equally involved in the development of the partnered service arrangements [10–12]. This was important for ensuring joint ownership and mutually beneficial outcomes. Nine papers described how active joint practitioner communication occurred, ideally in person and through regular clinical meetings, with information sharing and clinical problem solving, formulation of a common understanding about the nature of the partnership and its operation and a set of agreed goals. Active communication between practitioners included development of processes (e.g., care planning, guideline development, referral and follow up protocols) which enabled structured communication and communication channels such as regular meetings. This was helped by (and would help create) a receptive partnership culture, such as an "open door" communication style and a willingness to try out new ideas. A communication process also enabled a partnership to monitor and consider how it was operating to meet patients' needs [13–21].

In relation to goals, there was some evidence that patient involvement in care processes was also important. While only one paper specifically discussed the use of educational and behavioural change strategies to engage patients as collaborators with clinicians in setting their own goals and management of their care [18], many papers included these strategies in their intervention. Our reference group considered that patient set self-management goals could be a common focus for the different members of the clinical team. Hence the establishment of common goals as an aspect of clinical partnership formation may be enabled when the model of care includes patients as collaborators.

Joint development of a partnership through regular communication at the clinical level could assist the clarification of partner roles and attention to the different role concerns of the partners. Role clarity was the most frequently described aspect of partnership formation (14 papers), particularly as a barrier when this clarity was not present [16, 18, 19, 22–32]. We have extracted data from a few of these studies to describe the types of role conflict and also how role clarification occurred. Different or conflicting expectations about roles were reported as a barrier by Yaffe [25] and Macdonald [24]. Yaffe et al surveyed patients, family doctors and psychiatrists in a Canadian psychogeriatric outpatient clinic over 18 months. They found that clinicians disagreed about responsibility for treatment in 40% of cases, with family physicians expecting the psychiatrist to provide care after referral, while psychiatrists considered referrals were for assessment only. Furthermore, half of the patients said they did not know what to expect from their consultation with the psychiatrist. Macdonald et al interviewed 75 primary mental health workers in the UK and found that some had been expected to design their own roles without guidance. These workers reported feeling like "the meat in the sandwich" between primary care staff, who wanted them to adopt a direct clinical role, and mental health staff, who wanted them to take a consultation liaison role.

Team role clarification processes in the US IMPACT trial enabled the identification of safe practice boundaries [18]. The activities that supported role clarification were weekly team meetings between the care manager (Depression Care Specialist), the primary care physician and the psychiatrist as well as regular peer support telephone conferences with other care managers to discuss their work. Frazer et al found similar benefits in a survey of 13 Primary Care Graduate Mental Health Workers in the UK, but also reported that the development of an Integrated Care Protocol was a key component of role clarification [22]. A Canadian evaluation of primary care counsellors found that regular meetings for mutual support were important, as was supervision of the counsellor by a mental health specialist (in this case a psychiatrist) for case-based discussion and making treatment recommendations [12]. Kirchner reported similar findings about supervision in the US PRISM-E trial [17].

Another barrier to role clarity between primary care and mental health services concerned referrals, where there were different expectations about role responsibility for patients and dissatisfaction with referral processes [16, 27–32]. When primary care physicians referred large numbers of patients with "low severity" conditions to mental health teams, then this created an excessive demand on the team [28]. The team labelled many of these referrals as "inappropriate", often without considering the physician's needs or referral threshold [29, 30]. Chew-Graham and Slade reported on the failure of a validated Threshold Assessment Grid designed to aid referrals, as it was used by only 25% of physicians and few mental health teams [30, 32]. The clinicians found the Grid too simplistic and reductive. This was in a context without clear referral processes and tension about whether the mental health team role was crisis-response or caring for those with long term mental illness. Primary care physicians wanted direct access to psychiatrists and they felt that mental health team triage and the use of the Grid hindered access and undermined the doctor-to-doctor relationship [30, 31].

In addition to the supervision and mutual support of staff involved in linkage roles mentioned above, other factors were also described as supporting partnership formation. Five papers mentioned that an ideal attitude and skill set for staff was knowledge and skills in both mental health and primary care [14, 17, 18], a flexible work style that helped a person fit into the team [12, 17, 19] and a belief that collaboration is worthwhile [14]. However, one study found that difficulty in recruiting experienced mental health workers who were willing to take on collaboration roles was a barrier to collaborative care [33].

The physical features of a clinic were reported as an enabler, when the office of the care manager (Depression Care Specialist) was located in the primary care clinic for optimum team visibility and interaction [18], but a barrier when room space for the care manager (Primary Mental Health Link Worker) was inadequate [24]. Primary care physicians in larger practices of at least four physicians were found to be more amenable to collaborative care, made more referrals to community mental health teams and fewer referrals to in-patient and out-patient psychiatric services. These differences were attributed to consultation liaison arrangements and co-located psychology services in these larger clinics [26, 34]. Another study found, however, that the number of primary care physicians working in a clinic made no difference to their satisfaction in communication with psychiatrists [35].

A feature of papers that reported active joint practitioner communication processes was the inclusion of evaluation feedback in these processes. One of the largest trials in our review came to an explicit conclusion that mechanisms to feed back evidence about outcomes to partners was "the most important factor [for sustainability], cited in four of the five IMPACT study sites". These mechanisms included a reliable patient tracking, information and communication process to the team [26].

Organisational level

At the organisational level, five papers described the legitimisation of a collaborative approach through a supportive institutional environment of leadership and a change management for linked mental health services [17, 26, 33, 36, 37]. In the IMPACT study by Blasinsky et al, key informants indicated that strategic organisational leadership for collaborative care was important for the sustainability of the collaborative service model [26]. Our reference group also considered that the authority embodied in leadership can promote the legitimacy of collaborative practice. This may be important to counter resistance, such as that reported by Richards et al who found some concern from GPs that collaborative practice could challenge their leadership of patient primary care [33].

In the PRISM-E study, Kirchner et al compared leadership in two integrated care clinics, one that had successfully achieved integration and the other that had not [17]. The "unsuccessful" clinic had undergone change, which was perceived as chaotic and overwhelming. Part of this change included the employment of a new and influential leader who did not support integration, which subsequently led to "turf disputes" amongst the clinic services.

While we have proposed that partnership formation at the clinical level is necessary for the development of linked clinical services, this is unlikely to be sufficient to address organisation wide barriers, such as narrow and different service mandates and priorities that fail to address continuity of care, un-pooled resources and poor connections between senior managers across services. Discussing these wider organisational barriers, Rees et al concluded that organisational leadership is important to ensure accountability mechanisms are developed, to influence strategy and to provide resources for the change to collaborative practice [37].

Limitations

We did not conduct a quality assessment of the retrieved studies as they reported methods and results in different ways, with different theoretical perspectives and varying levels of detail. This range of studies in our broad review (over the two parts) would have made applying comparative quality criteria complex and outside our funded timeframe of one year. This is a limitation of our findings, as we cannot give weight to these findings beyond an estimate of importance of each factor based on the number of studies in which they were reported.

While the findings from this review are informative, about the factors that enable the formation of mental health service partnerships in primary care, the multifactoral and context dependent nature of these factors does limit the generalisation of the findings. This would particularly be the case when considering whether findings from one country apply to another.