The search strategy identified 1870 references (Figure 1). After removal of duplicates 1802 abstracts were examined for relevance and 68 full text references were obtained for full text screening. Hand-searching of references lists of included articles yielded an additional three articles. In total 71 articles were assessed for eligibility, of which 13 articles were selected for data extraction and analysis.
Fifty-eight studies assessed for eligibility were excluded. Eighteen were grey literature, 17 did not include the primary health care nurse, six were poster abstracts and the studies not published, and 17 papers were removed as they were multiple publications reporting on the same intervention and were included as supplementary papers. Three were duplicate studies [19-21] and two studies [22,23] were excluded as the outcomes had not been published. The authors of these studies were contacted. Bryans et al.,  did not publish the outcomes of a survey study on primary health care nurses and dementia care due to significant loss to follow-up. For similar reasons, Perry et al.,  did not publish the outcomes of the dementia training programme on diagnostic assessment and management of dementia by primary care nurses.
Of the 13 included studies, three were RCTs, three utilised survey questionnaires, four were mixed-method studies with one of these a Rand-Delphi study, and three were qualitative studies using interviews.
The studies were conducted in the Netherlands (n=1), Germany (n=1), United States of America (n=1), The United Kingdom (n=5), Australia (n=4) and one was conducted across the Netherlands and the United Kingdom (n=1).
Due to the heterogeneity of the studies a pragmatic approach to reporting the findings of was taken by aggregating studies by methodology. The heterogeneity did provide important insights into the variety of involvement of primary care nurses in delivering dementia care.
Four studies [21,24-26] evaluated dementia care management in primary health care. Exploring dementia care knowledge and attitudes of primary health care practitioners was the focus of three studies [3,11,12]. Two studies [27,28] explored participant experiences of dementia care delivery in primary health care and one study  explored service use and reported unmet needs of people with dementia and support person(s). Investigating the implications of early recognition of dementia for the roles of the primary health care team was the focus of one study  The authors of one study  developed quality indicators for dementia care in primary health care settings and one study investigated the value and useability of an online dementia management tool for health professionals . The study interventions and outcomes are described in Tables 5 - 8.
Randomised Controlled Trials
Three studies utilised an RCT [21,25,26] to investigate the impact of collaborative care on quality of life for people with dementia and their caregivers. The study by Van den Dungen et al.,  also included an evaluation of family practitioner training on diagnosis of mild cognitive impairment.
In all three models of care the nurse was the care manager who worked in collaboration with the primary care doctor. All care management models followed a structured assessment and care planning protocol. Care management ranged in duration from six  to twelve months [21, 25]. In two studies [25,26] the care managers were registered nurses, with Van den Dungen et al.,  specifying the nurse as a primary care nurse who acted as the study nurse. In the third study  the care manager was a geriatric nurse practitioner. All the nurses received dementia specific training and were integrated into the primary care team with only one care manager providing the dementia care management within the patients’ home . In addition to training, in the model of care described in Callahan et al.,  the nurse received weekly support from a geriatrician, geriatric psychiatrist and a psychologist.
Callahan et al.,  and Thyrian et al.,  reported a significant decrease in behavioural and psychological symptoms of dementia and caregiver stress with dementia care management, however, Thyrian et al.,  reported there was no significant improvement in quality of life overall. Despite reporting that dementia care management had no impact on quality of life measures for patients or their care-givers, Van den Dungen et al., recommend that collaborative care with nurses in primary care deserves further exploration.
Survey Questionnaire studies
Three studies reported survey results [11,12,29]. Manthorpe et al.,  and Trickey et al.,  investigated dementia knowledge and attitudes of community nurses (CN), health visitors, community mental health care nurses (CMHN) and PNs in the provision of care for people living with dementia. The third study  explored service use and unmet needs of people with dementia recruited a decade apart.
Manthorpe et al.,  reported all groups of primary health care nurses had similar knowledge related to the early signs and symptoms of dementia. However, PNs were less confident in providing advice and support than CMHNs. In the study undertaken by Trickey et al., , PNs completing the Over-75 year health check were less likely than other nurse groups to take any action, other than to refer to the GP, when presented with a person living with dementia and their support person. The Over-75 year health check is an annual health check including a mental assessment for people aged over 75 years .
Gilbert et al.,  reported that support person(s) were increasingly contacting a PN for support with less evident use of CNs, health visitors and CMHNs. This may in part be attributed to greater access to a PN and the changing nature of the PN role with an increased focus on chronic disease management. Support person(s) reported that they were still not getting the advice and support they needed.
Authors of all three studies identified a need to improve PN knowledge of dementia and its management. In the study by Trickey et al.,  participants reported guidelines would be helpful to address gaps in knowledge and to standardise practice.
Four studies reported mixed-methods research results [3,24,31,32].
Perry et al.,  used a RAND modified Delphi method to construct a set of quality indicators for dementia diagnosis and management in primary care in the Netherlands. PNs were involved in the selection and validation process of the quality indicators. Of the final 23 quality indicators, two explicitly describe collaboration between the GP and the PN, an area in which the authors suggest improvement is highly recommended. A further three quality indicators emphasise the importance of developing and reviewing individualised care plans. This is commonly a PN role that is established and accepted in primary care settings . Millard et al.,  explored dementia literacy in a general practice setting. In this study two-thirds of the PNs reported a lack of dementia training. Despite this self-perceived lack of training, three-quarters of the PNs reported that the primary care doctor or nurse was the appropriate person to discuss dementia with patients. Ollerenshaw et al.,  suggest that PNs may find an on-line dementia management support tool useful. Iliffe et al.,  adapted a US model of primary care based case management (PREVENT) for people with dementia and tested its implementation in UK general practice. Despite case managers, patients and support person(s) reporting a positive experience and perceiving benefits of case management, Iliffe et al.,  suggest that case management does not fit easily into practice routines and that it was not substantially beneficial for patients and support person(s).
All three qualitative studies [27,28,30] used interviews to explore experiences of primary health care practitioners, patients and support person(s), of dementia care. Dodd et al.,  used semi-structured face-to-face interviews to contrast study participants’ experiences of a new primary care led dementia service with existing secondary care based memory services in Bristol, UK. Dodd et al.,  used a semi-structured face-to-face interview to investigate participant’s experiences of a new primary care led dementia service in South Gloustershire, UK. In both these studies [27,28] the nurses were seconded from secondary care dementia services, with each nurse working with a group of primary health care clinics. Patients and support person(s) reported primary care led services to be positive and there was uniform praise for the work by the memory nurse. GPs reported they valued the advisory role provided by the memory nurse. Manthorpe et al.,  explored implications of the early recognition of dementia for inter-professional working using focus group interviews. In this study the PN was identified as the practitioner most appropriate to take on screening for dementia and monitoring, however community mental health care nurses were considered to have the skills and capacity to take on long-term and complex cases.
Risk of bias
The methodological quality varied across the studies (Tables 1 -4). The qualitative studies and all but one of the mixed methods studies rated high according to the quality appraisal criteria. Two of the three RCT studies lacked allocation concealment, blinding and presented incomplete outcome data which compromised their quality. The survey studies were of mixed quality with two of the three studies introducing selection bias and no sample size was based on consideration of statistical power.
In addition to these limitations, Callahan et al.,  describe their study as unable to identify which of the subcomponents of the intervention were most effective in achieving the outcomes. Van den Dungen et al.,  reported the rates of MCI or dementia identified were lower than expected. The authors state the reasons for this may have included a type 2 error with a low sensitivity of the cognitive tests performed by PN. In addition, there was sub-optimal implementation of the intervention with the family practitioner not always performing further diagnostic assessments on all persons referred by the PN . Thyrian et al.,  describe limitations of the study including potential selection bias as screening and recruitment were part of routine care. The intervention and control groups had an uneven number of participants; the GPs in the control group had fewer patients. In addition, the GPs may have become aware of their assignment to the control or intervention group .
Trickey et al.,  describe a methodological limitation of using a vignette that may more correctly explore current practice rather than knowledge and attitudes . Iliffe et al.,  report time constraints for the case management role of the PNs may have meant there was insufficient time to show the potential of case management.