Stroke covers several processes that cause permanent damage to a part of the brain by obstruction to cerebral blood flow or intracranial haemorrhage. This causes impaired physical function, including paralysis, loss or distortion of higher-order functions like language, memory and personality, and alteration of automatic functions like swallowing, speech, and continence. Significant stroke leads to major disability. Stroke is the third largest cause of death in Australia. In 1997, the annual cost of stroke care was $A555m, with the average cost of care twelve months after a first time stroke being $A18,956, and the lifetime cost being $A44,000 per case. Stroke care accounted for around 269,000 GP consultations per annum in Australia 1.

Coordinated care approaches, which aim to provide comprehensive care specific to the needs of the individual patient or patient group, are presumed to offer benefits to those with complex needs such as those recovering from stroke. Different approaches include case management, disease management programs or integrated service delivery 234. Multidisciplinary care planning is one of these approaches.

Because of the range and seriousness of the impacts of stroke on patients and their families, multidisciplinary care is frequently required5. Such care enables participation by health professionals from different disciplines, services or sectors in planning and/or delivering care. However, given structural differences in the way hospitals and community agencies work, implementing multidisciplinary care across the primary/secondary divide can be exceedingly difficult to achieve.

The aim of this systematic review is to assess the impact of co-ordinated multi-disciplinary care planning involving primary care professionals, either wholly within primary care or by primary-secondary care teams, on outcomes in stroke, relative to usual care.

The approach to the investigation of the review was derived from the work of Mays, Pope and Popay 6 in describing frameworks for the review and synthesis of health services evidence which have to deal with different study designs, research traditions, theoretical orientations, and disciplines. Interventions of this nature are complex and context dependent, creating challenges in identifying and comparing studies. Studies are rarely replicated exactly. Analysis is question-driven, enabling examination of underlying context issues and background assumptions as well as the outcomes of a study.

This approach is not dissimilar in intent to other approaches to reviews of complex interventions such as Medical Research Council's Framework 7 where the emphasis is on the process relating to developing and evaluating a complex intervention or Theory of Change 8 which looks at the function and purpose of the intervention rather than the compositional elements alone. As it may be impossible to apply the method of one study to another context completely, these approaches enable systematic reviewing of interventions which may have multiple components, multiple outcomes and different implementation strategies.

As health service interventions are multifaceted and are variously affected by the presence or absence of an evidence-base, methods used, stakeholders and contextual factors, to appreciate the impact of multidisciplinary care planning interventions better, we not only examined effectiveness by reviewing controlled trials, but also extended the scope to capture participants' views on how multidisciplinary care operates, and clinical experience (by reviewing qualitative studies, editorial opinion, and grey literature reports, case studies).

Evidence-based guidelines were examined to identify how multi-disciplinary care planning was best conducted, who should be involved and what roles participants should play.

Given the heterogeneity of study types, models of care planning and team organisation and outcome measures, a narrative approach to review and analysis of extracted data was undertaken.

The use of a systematic review to describe the outcomes of a complex intervention has to take into account the context in which the study was performed 627. This review has identified various models of care planning and coordination that have been utilised in stroke management in the primary care setting. National health systems use an array funding and administrative structures which support models of care provision. However, two basic coordination approaches emerge. The first assumes that clinical decision making is a specialist task, and the primary care role is one of facilitating the implementation of the care plan and timely identification of complications. The second overarching model assumes transfer of care and responsibility to the primary care team, and use techniques that engage actively with primary care providers with transfer of care in mind. Prevention of secondary stroke as a primary care responsibility is common to both models identified and is acknowledged in stroke guidelines 28.

It is uncertain whether multidisciplinary care involving GPs improves outcomes in patients with completed stroke. Interpretation of the results is difficult, as results of the two largest studies appear contradictory, and analysis is complicated by the diversity of outcome measures. The Trondheim study tested the hypotheses that close cooperation between specialist team and primary care team was beneficial for the patient's recovery, and that rehabilitation conducted at home was more beneficial than rehabilitation unit based care. This was the study that improved the proportion of patients achieving independence (with a number needed to treat of seven) and modest improvement in quality of life at twelve months 212930. By contrast, the other large study aimed to compare the effectiveness of conventional care with specialist stroke unit care including intensive discharge planning: this trial had limited involvement of primary care professionals 19. It showed no differences between intervention and control groups. The smallest study applied to rural patients 20: there were also limitations to the degree of integration of specialist and primary care services. It too showed no differences in outcomes.

Secondary care providers may make assumptions about what should be happening at a primary care level without systems to check the presumed care actually happens. Particularly when there is patient follow-up in the community by a secondary service, primary care service providers may not know what their role is expected to be, without dialogue and negotiation.

The question thus arises whether systematic effort put into engaging GPs and other primary care providers is effort well spent. While it is difficult to draw conclusions, there may be at least process benefits. It makes sense to ensure that primary and secondary care providers know what their respective responsibilities for the care of the patient are. Without this, some essential tasks assumed to have been done by another professional may be missed, while other tasks are duplicated.

Education is one avenue to increase GP's knowledge of stroke care requirements, but outcomes rely to some extent on the GP's own areas of interest 31 and mode of delivery 3233. Case-based education relating to problems experienced by current patients is effective in improving knowledge and practice. Engaging primary care providers including GPs in multidisciplinary discharge planning thus provides excellent opportunities to use the index case to teach.

Implementation of change in care approaches seems to depend on professionals working collaboratively. Finding mechanisms and forms of collaboration that support patient outcomes is an increasing important area of implementation research. The Alberta Primary health care project report identified eight critical factors for forming successful teams including "involving all members in planning and coordination" and "defining roles and responsibilities" 34 Recent work suggests that personal relationships between specialists and primary care providers are essential for improved communication 35.

There are limitations to this study. Frameworks for the review and evaluation of health services and care programs are still developing. The interventions and outcomes of the included studies in this study are heterogeneous, and therefore the review presents the findings in narrative form only. Although a comprehensive search strategy was undertaken, given the complexity of possible sources, terminology and brevity of descriptions of interventions, some literature may have been missed. As there is no single definition of the tasks and composition of multidisciplinary care planning, inappropriate inclusion or exclusion of a study may have occurred. Dual review for inclusion sought to address this issue. Although all included studies had multidisciplinary care planning as part of the interventions assessed, different processes and arrangements meant that it was not possible to determine the degree to which multidisciplinary care planning as an independent element led to the observed outcomes.

It is unclear whether coordinated care planning involving GPs and primary care health professionals makes an unequivocal difference to patients outcomes to patients with completed stroke. The inconsistency in results suggests that further study around the type of involvement of GPs in the multidisciplinary planning process could be illuminating.

MeSH: Medical Search Headings; GP: General Practitioner; Wks: weeks; Mths: months; pa: per annum; CI: Confidence Interval; NS: Not significant; OR: Odds Ratio.

The authors declare that they have no competing interests.

GM conceived this systematic review, devised the search strategy, provided the initial analysis of the data and coordinated writing of the article. RB conducted the literature searches, and with GM assessed which articles should be included. LE searched the grey literature and provided advice on the search strategy. JT managed the project of which this review was a part, provided conceptual advice on the conduct of the project, and provided extensive editorial input into the final manuscript. All authors provided input into the interpretation of the data, and provided editorial input into the final manuscript.