Outpatient services and primary care: scoping review, substudies and international comparisons

Search strategy

We reran the previous search strategy, adding search terms for the new categories. The literature was searched from February 2005 to April 2014, starting with the end date for each of the previous searches.

The databases searched in the current review were:

• MEDLINE^® (via Ovid) (February 2005 to April 2014)

• EMBASE (via Ovid) (February 2005 to April 2014)

• Health Management and Information Consortium Health Management and Policy database (via Ovid) (February 2005 to April 2014)

• The King’s Fund database of grey literature (http://kingsfund.koha-ptfs.eu/) (February 2005 to April 2014).

The previous review included a search in the Cochrane Database of Systematic Reviews and a search in the Campbell Collaboration Library of Systematic Reviews. We did not replicate those searches in this update, as we found that references from those databases were picked up in the other searches.

The detailed search strategies can be found in Appendix 1.

We became aware that many primary care trusts (PCTs) have produced guidelines to reducing referrals, often drawing on common sources such as the NHS Institute and The King’s Fund. These were not included in the review as, on investigation, we discovered that they reiterated advice on managing referrals without any evaluative component and did not, therefore, meet our inclusion criteria (see the following section).

Inclusion and exclusion criteria

Surgical clinics

Minor surgery principally became of interest because the 1990 GP contract included financial incentives for GPs to carry out minor surgery in their own practices.

Nine studies (10 papers) were included in our previous review, from which we concluded that ‘some evidence suggests that the quality of care provided in general practice was initially poor due to inadequacies in GP training, problems in maintaining surgical skills given the low patient volume, and inadequacies in the equipment and/or procedures used to sterilise surgical implements’. We also suggested that ‘many of the additional patients receiving minor surgery under the conditions of the 1990 contract may not have previously been referred to hospital, and that GPs may have used minor surgery in place of cheaper, equally effective treatments’.

Since then, the policy context has changed in that the financial incentives for GPs to carry out minor surgery have reduced. However, some general practitioners with a special interest (GPwSIs) are contracted specifically to carry out minor surgery.

An additional six papers were found in our updated scoping search (Table 2). 10–15

One study was a randomised controlled trial (RCT) which concluded that patient satisfaction was high, access to care was improved and costs were lower when procedures were carried out in primary care (the mean cost for hospital-based minor surgery was £1222 and for primary care was £449). However, the quality of minor surgery carried out in general practice was not as high as the quality of that carried out in hospital. 12 This related in particular to the incomplete excision of skin cancers. Using completeness of excision of malignancy as an outcome, hospital minor surgery was more cost-effective than surgery in primary care. However, another study of highly trained GPs (GPwSIs) found no difference in the incomplete excision rate between GPwSIs and hospital doctors. 16

Studies in the past have suggested that, with adequate training, GPs may safely perform procedures such as hernia repairs11 and endometrial thermal ablation14 in primary care.

There is little additional information on the cost-effectiveness of minor surgery in general practice. Van Dijk et al. 15 found that Dutch practices which carried our minor surgery referred fewer patients with those conditions to hospital, but did not carry out a formal cost-effectiveness analysis. In practices in rural USA, Nelson et al. 13 found that it was cost-effective to train primary care doctors to carry out knee injections for patients with osteoarthritis.

Medical clinics

In our previous scoping review, we concluded that ‘there was a marked dearth of research comparing general practice care with hospital care’ for major chronic diseases. However, we concluded that ‘If care is well structured – there is a disease register and recall system, with clinical reviews conducted in accordance with evidence-based guidelines – then short-term health outcomes for patients appear to be as good as those achieved in hospital outpatient clinics’. However, there was insufficient evidence to draw conclusions about costs of primary versus secondary care of chronic conditions.

The policy context of the earlier review was the 2004 GP contract which included major financial incentives for GPs to monitor chronic conditions in general practice [the Quality and Outcomes Framework (QOF)]. These changes have become firmly embedded in the structure and practice of primary care since then.

An additional 16 papers were identified in our updated scoping search (Table 3). 18–33

The additional studies we found confirm that high-quality care can be provided for patients in primary care, including for diabetes,21,25,32 adult asthma,31 childhood asthma,26 colonoscopy,29,30 alcohol dependence22 and chronic obstructive pulmonary disease (COPD). 33 Providing additional resources in primary care can also reduce secondary care utilisation, for example through the use of a nurse-/pharmacist-led clinic for patients with chronic pain. 20

In some studies, the primary care alternative to specialist referral is a different and less intensive approach to management. For example, van Boeijen et al. 19 found that two management approaches in primary care (guided self-help and structured guidelines) were as effective for patients with anxiety as referral to a psychologist for cognitive–behavioural therapy. This study emphasises that, although a common assumption is made that the same care will be provided whether in a primary or a secondary care setting, this may not in fact be the case, either by design or by default.

The support and training for primary care staff is important. For example, in the case of urological conditions, GPs did not feel that they had the skills to take on the additional roles,27 and primary care nurses in primary care did not feel confident in dealing with some of the skin conditions they were managing. 23 However, the additional studies found in this review support the earlier conclusion that a wide range of conditions can be managed effectively in primary care if adequate training and support is given to staff.

However, there are few studies of the cost-effectiveness of transferring chronic disease management from secondary to primary care. The potential for cost saving by moving services from specialist to primary care was emphasised in a modelling study by Berger et al. ,18 who found that even small increases in the number of patients with neuropathic pain managed in primary rather than secondary care in Norway could result in substantial overall cost savings. Economic analyses depend in part on whether care improves (or gets worse) as a result of transfer to primary care. Two reports of the same study, Dusheiko et al. 24 and Martin et al. ,28 suggested that, for stroke care, improvements in management since the 2004 GP contract had led to a reduction in hospital expenditure (significant for both outpatients and admissions), but they did not find evidence of this for other conditions they studied. 24,28 Several studies report lower rates of secondary care utilisation when care is moved from secondary to primary care but there is a striking absence of studies which estimate the impact of transferring care on the overall health budget (i.e. primary and secondary care combined). For example, in a study of colonoscopy in general practice,29 the cost of the procedure was lower in primary care, but the authors did not report a full economic analysis taking into account the total number of referrals for colonoscopy to hospital and primary care-based clinics combined or referrals from the primary care clinic to secondary care.

General practitioners with a special interest

General practitioners with a special interest were developed in the early 2000s principally to increase the capacity of the NHS to provide specialist advice and to reduce outpatient waiting times. In our previous review, we concluded that GPwSIs provided care that was of an equivalent standard to that provided in hospital outpatient clinics, although systems for monitoring quality and outcomes varied, with data on long-term follow-up of patients largely absent. GPwSIs appeared to be costlier than outpatient clinics in part because the rates of pay for GPwSIs are higher than for non-consultant doctors in outpatient clinics. However, we commented that the costs were highly likely to be context specific. It should also be noted that GPwSIs were introduced without there necessarily being an intention to reduce cost, as a main aim was to increase capacity to reduce outpatient waiting times.

We identified 10 additional studies and relevant commentaries in this updated scoping review (Table 4). 10,34–42

Salisbury et al. 41 randomised dermatology patients to being seen either in a hospital outpatient clinic or by a GPwSI in primary care. They found no difference in clinical outcomes, but waiting times for the GPwSI were shorter and patients preferred being seen in the primary care setting. However, an economic analysis suggested that referrals randomised to the GPwSI were 76% more expensive than the hospital equivalent (£208 vs. £118). 35 In a non-randomised study of a GPwSI for headache compared with a neurology outpatient appointment, there were again no differences in clinical outcomes and patients again preferred the GPwSI setting. However, in this case the costs of the GPwSI appointment were lower than those for the hospital clinic. 40

One study randomised patients to seeing GPwSIs in either a hospital or a practice setting. 34 There were no differences in clinical outcomes, but patients again preferred being seen in primary care settings.

Overall, it is clear that appropriately trained and supported GPwSIs can provide a high-quality service that is valued by patients, and patients generally prefer to be seen in community settings. GPwSIs are now operating in a wide range of specialty areas37 and there is clearly significant potential for patients currently seen in outpatient clinics to be seen by GPwSIs; for example, Gilbert et al. 36 estimated that 23% of patients seen in a chest clinic could be seen by a GPwSI.

When interpreting the impact of GPwSIs on waiting times and costs from studies in the early 1990s, it is important to appreciate that these clinics were generally introduced as additional services designed to reduce waiting times at a time of major investment in the NHS. Now that the NHS faces reduced additional investment, GPwSI services are more likely to be intended as a substitute for hospital care. However, it does not necessarily follow that the introduction of GPwSIs will reduce demand on hospital services, although they have the potential to do so; the provision of additional services could result in ‘supply-induced demand’ if GPs’ referral thresholds change. For example, in an earlier study described in our previous review, Nocon et al. 39 reported that although outpatient attendances reduced following the introduction of GPwSIs, overall attendances rose. Indeed, we found one report in which hospital attendances increased after the introduction of GPwSIs in the community. 38 In an evaluation of practitioners with special interests among a range of ‘closer to home’ demonstration sites set up by the Department of Health in England in the mid-2000s, Sibbald et al. 42 suggested that the difference in cost per patient for commissioners may be lower for GPwSIs than hospital care national tariffs but that this might be explained by GPwSIs seeing less complex cases. They similarly cautioned that GPwSIs may increase demand for outpatient services.

It is, therefore, not clear whether or not GPwSIs provide a cost-effective alternative to outpatient clinics. In addition to the potential for ‘supply-induced demand’ referred to in the previous paragraph, there are two reasons why GPwSI clinics could be more expensive. The first is that hospital clinics are staffed by a mixture of consultants and cheaper subconsultant grades, whereas GPwSIs are paid at the equivalent of consultant grade. There are also economies of scale in hospitals which may allow consultants working in a hospital clinic to give opinions on a larger number of patients when they are supported by junior staff.

Discharge from outpatients to primary care

In our previous review, we found that both patient-initiated outpatient follow-up and transfer of follow-up to primary care are plausible strategies for reducing outpatient attendance rates and overall NHS costs without adverse effects on the quality of care or health outcomes. Patients in general found GP visits more convenient, less time-consuming and less expensive than outpatient attendance. However, the acceptability of alternative discharge arrangements to patients, specialists and GPs was variable and far from universal.

An additional 13 studies43–55 including one systematic review54 were identified in our updated search (Table 5).

Five of the 13 studies involved cancer patients. 45,46,49,50,55 In one RCT in Canada, there was no difference in outcome at follow-up (for a mean of 3.5 years) when patients (968 women with breast cancer) were seen in an oncology clinic compared with a family practice. 45 Likewise, in a trial in Norway there were no significant differences for a range of outcomes in 110 patients with colon cancer randomised to specialist or GP follow-up, which included a decision support tool for patients and GPs. However, costs were reduced in the GP follow-up arm. 43 The acceptability of follow-up in primary care was assessed by Hall et al. 46 in a qualitative interview study. They found that patients would generally be willing to have GPs share their cancer follow-up, with the caveat that the GPs had received extra training and were appropriately supported by secondary care specialists. GPs in the study stressed the importance of maintaining their own clinical skills and having reliable clinical and administrative support from secondary care. In a modelling study, Lu et al. 49 showed the considerable cost savings that could accrue from transferring follow-up of cancer patients to primary care, with no significant difference in patient outcomes. With the limited evidence on cancer follow-up from our previous review, there is clearly potential for the routine monitoring of patients who have had cancer to be carried out in primary care, provided that clear protocols and training are available to primary care physicians.

Two other studies looked at the follow-up of people, the first following a diagnosis of dementia51 and the second a limited study of patients with renal failure,52 both concluding that patients could be discharged to primary care without adverse consequences. One-third of patients with recurrent anogenital herpes48 could also be successfully be transferred to care from their GPs, although a proportion of patients did not want to be transferred as they did not want their GP to know about their diagnosis.

The systematic review54 related to telephone consultations by the specialist (surgeon or specialist nurse) as an alternative to outpatient clinic attendance following surgery. Because of the poor methodological quality of the studies included, the authors felt unable to draw any firm conclusions about the role of telephone follow-up in this situation.

The additional studies found in this scoping review, although limited in number, support the ability of GPs to follow up patients across a range of diagnostic groups as an alternative to hospital follow-up. Although many of the procedures and investigations required during follow-up may be available to GPs, general practices are not currently as well organised for ongoing follow-up of many conditions as they are, say, for the routine monitoring of diabetes or asthma. Therefore, if patients requiring ongoing follow-up are to be discharged back to primary care, it is important to ensure that general practices have the administrative support and resources to ensure that follow-up protocols can be reliably followed, and that there is adequate support from specialists when queries arise or problems occur. One option is to involve patients more closely in the decision. For example, in the studies by Augestad et al. 43 of colon cancer and Lund et al. 50 of prostate cancer, leaflets were given to the patients explaining what sort of follow-up they should expect from their GP and when to expect particular tests, etc. This information was also provided in leaflets for GPs.

Where the need for follow-up relates to patients’ symptoms, the patient may be able to assess the need for hospital follow up. For example, in a 6-year randomised trial of patients with rheumatoid arthritis47 patients were discharged and allowed to make follow-up appointments when they felt they needed one, rather than being given routine appointments. All outcomes (clinical, patient experience and economic) were in favour of patient-directed follow-up.

The availability of electronic records shared between hospitals and GPs could increase the availability of decision support when patients are no longer seen regularly in the specialist clinic. For example, Hennessey et al. introduced an electronic decision support system into their clinic which decided when prostate cancer patients needed to have their next outpatient appointment and/or prostate-specific antigen (PSA) test. 55 This was based on the computer’s assessment of a PSA test which had been carried out by their GP. Patients were not formally discharged from the clinic, but the authors reported that the number of outpatient attendances was greatly reduced.

One question which this review does not answer is how many patients are being followed up in hospital clinics who could be discharged to primary care. In a study in Birmingham, UK, GPs were given a financial incentive among other things to review the records of patients being followed up in hospital diabetic clinics and discharge those who did not meet specified referral criteria. A substantial reduction in both new and follow-up referrals was reported (odds ratios 0.69 and 0.77, respectively), although there were no data for comparators in this simple before-and-after analysis. 44

Direct access to diagnostic tests and investigations

In our previous study, we described studies evaluating GPs’ access to electrocardiograms (ECGs), echocardiography, ultrasound, upper gastrointestinal endoscopy, sigmoidoscopy and various types of radiology, including five studies of direct access to magnetic resonance imaging (MRI) or computerised tomography (CT). We concluded that GPs having direct access to these tests was popular with patients and that a significant proportion of investigations (varying between studies and types of investigation) were thought to save an outpatient referral. Although there were differences in the types of patients referred, the diagnostic yield was similar when comparing tests ordered by GPs and specialists. We found few data to assess whether or not direct access to investigations led to a net reduction in NHS costs (i.e. whether or not the increased costs of testing in primary care were offset by reduced costs of attendance at outpatient clinics).

Since our first review was published, GPs have gained more or less universal direct access to ECGs, echocardiography, ultrasound, upper gastrointestinal endoscopy and, in places, MRI. Sometimes access to these investigations requires a range of criteria to be met, usually contained in a referral pro forma.

In this scoping review, we identified 25 additional papers (Table 6). 56–80 These included direct access to MRI (three studies),63,64,72 CT (three studies),57,70,74 diagnosis and management of deep-vein thrombosis (DVT) (four studies),59,60,73,75 retinal photography for diabetic retinopathy (six studies),56,61,62,67–69 ultrasound (one study)79 and gynaecological ultrasound (two studies),65,78 siascopy for suspicious moles (one study),76 respiratory tests (two studies),66,80 lung cancer diagnostic (one study),58 cardiac arrhythmia monitoring (one study)71 and computer-aided cardiac auscultation (one study). 77

Direct access to services

In our previous review, we concluded that GPs generally refer appropriately to direct-access services, but that there was inconsistent evidence on the impact on the overall demand for services, with savings in hospital costs sometimes offset by an overall increase in demand.

An additional seven studies81–87 were identified in our updated scoping search (Table 7).

For some services, there appears to be a clear-cut benefit in streamlining a manifestly inefficient referral pathway. For example, patients for whom a hearing aid is needed can be referred directly to an audiology clinic without first being seen by an ear, nose and throat (ENT) surgeon. 82 The audiologist is able to pick up those patients requiring assessment by a specialist (in this case, 16%) and it is unlikely that many patients are prescribed hearing aids unnecessarily, making this a highly cost-effective alternative to consultant referral.

Referrals to a service may increase when direct access is provided, and then there is a question of whether or not this addresses an unmet need. For example, the provision of a direct-access low-vision aid service in Wales increased the number of low-vision aid assessments by 51% and identified what the authors described as a ‘considerable unmet burden of need’. 85 There were major reductions in waiting times, significant reductions in visual disability in patients referred and very high patient satisfaction.

Musculoskeletal services are more complex. Many areas have introduced referral pathways to reduce referrals to orthopaedic clinics. These pathways may include direct access to physiotherapy (face to face or by telephone), osteopathy, chiropractic and acupuncture. The difficulty in evaluating these services is that referrals are often for conditions which are themselves self-limiting. For example, RCTs88,89 suggest that the effect of manual therapies for back pain is generally small and shortens the duration of disability rather than curing patients who would otherwise have remained disabled (although one RCT in this review suggested a long-lasting effect of direct access to acupuncture for low-back pain86). In other reports, which indicate high patient satisfaction and improvements in symptoms following direct access to new services (e.g. Bernstein81 and Gurden et al. 83), it is hard to attribute improvements in patients’ symptoms to the new service in the absence of any comparison group.

The issue the NHS faces with musculoskeletal problems is a huge burden of morbidity and a limited specialist resource. Where the patient requires physiotherapy and the only route to this is to see an orthopaedic surgeon (which is no longer commonly the case), is it clearly more efficient to provide direct access to physiotherapy for those patients. However, the provision of direct access also increases demand; for example, a study described in our earlier review showed that when direct access to a musculoskeletal service was introduced in Wales, referrals more than doubled. 84

We found other studies recommending direct access to services but with little comparative evaluation. For example, Julian et al. 87 found that women with menstrual problems could be effectively managed by GPs using a protocol which allowed access to investigations and direct listing for surgery, with no significant differences in patient outcomes compared with a consultant-led menstrual clinic. Studies of direct access to services generally show high patient satisfaction and a decrease in waiting times. However, almost all of these services were introduced at a time of major increases in NHS funding and top-down initiatives to reduce outpatient waiting times between 2000 and 2010. The additional services were often funded in addition to existing outpatient services, with this increase in supply generating its own additional demand. It therefore remains very hard to comment in general on the cost-effectiveness of direct access to services.

Perhaps the most promising change during this period, but one which was addressed only tangentially in the studies we found, was the increasing tendency for locally agreed pathways between GPs and specialists to be used as the basis for access to services (e.g. criteria which need to be met in order for a patient with back pain to be referred directly for MRI). We note also that during this period (though maybe less widespread), specialists have been introducing similar types of criteria for outpatient referrals also. These are often articulated in electronic referral forms which require certain criteria to be met (or tests carried out) before the referral is accepted. We discuss these further in Professional behaviour change, Guidelines, including referral pro formas.

Shifted outpatient clinics

Shifted outpatient clinics involve hospital specialists visiting premises outside the hospital site to provide care. Instead of travelling to hospital clinics, patients visit specialists in these community settings. In our previous review we found that these clinics were popular with patients. However, specialists generally see fewer patients in a community clinic (partly because they are less likely to be working alongside junior staff) and a significant proportion of patients seen in community-based clinics needed to be seen again in the hospital, and community clinics may, therefore, be more expensive than hospital outpatient clinics. A benefit expected of some clinics was that interaction between specialists and GPs would help to educate GPs, but this did not generally occur. A Cochrane review included in the previous review drew similar conclusions. 90

The four additional studies identified in this review did not, in general, add to these conclusions (Table 8). 39,42,91,92

In most cases, the community-based specialist service was provided as an addition rather than an alternative to outpatient clinics: patient satisfaction was high and waiting times were reduced. However, it was not possible to determine whether or not reductions in waiting time were simply as a result of increased specialist capacity, or what effect a new locally based service had on demand. A study by Nocon et al. 39 showed that when a community-based diabetes service was introduced (some run by specialists, some run by specialist nurses and some run by GPs), referrals to hospital outpatients reduced, though overall the number of referrals rose.

Issues which are clear in relation to community-based specialist clinics are:

• They are popular with patients and often associated with reduced travelling time.

• They may reduce waiting times when they are part of an expansion of specialist capacity.

• The ready availability of specialist opinion in local clinics may increase referrals, with uncertain benefits to patients.

Key issues which remain unanswered in relation to community-based consultant clinics include:

• Whether or not community-based clinics can ever be cost-effective from the perspective of the NHS, bearing in mind the smaller number of patients generally seen in community clinics. The reduced number of patients seen in community clinics results in part because specialists are often accompanied (or even outnumbered) by junior staff in hospitals, enabling the consultant to supervise the management of a larger number of patients.

• How community clinics can be used to improve interaction with GPs, potentially leading to the upskilling of GPs and fewer referrals in future. There is little evidence that this potential benefit of moving consultants into the community has been realised to date.

In our view, current policies to move consultant clinics into the community can be justified only if (a) high value is given to patient convenience in relation to NHS costs or (b) community clinics can run at a scale to enable the efficiencies of patient throughput in a hospital clinic to be realised.

Alternative models may, however, be used to bring specialist expertise into primary care. For example, Wiedemer et al. 92 described a nurse practitioner and a pharmacist who ran a clinic for patients with chronic pain on opiates. There was no evaluation of the impact of the scheme on specialist referral outside the practice, but management by primary care physicians improved over time (e.g. greater use of treatment agreements). We describe in the other sections models in which non-medical specialists may contribute expertise to primary care. Sibbald et al. 42 also looked at a range of relocation services, where the venue of specialist care shifted from hospital to the community and, in some cases, the consultants provided community-based outpatient clinics with GPs or other practitioners with a special interest. However, in their evaluation of these services, the impact on interaction between consultants and GPs or upskilling in community was still not clear. The impacts on waiting time and satisfaction were in line with other studies.

Specialist attachment to primary care teams

Although bearing some similarities to shifted outpatient clinics, this group of studies involves active participation of both specialist and generalist in the clinic, and hence includes elements (particularly educational elements) which are absent from the shifted outpatient model. In our previous review, we included practice-based physiotherapy in the category (and concluded that primary care-based physiotherapy services were more cost-effective than hospital-based physiotherapy). We have not included physiotherapy as a ‘specialist resource’ in this review as physiotherapy has, since our first review, become an established part of the primary care environment. In our previous review, we also included CMHTs in this section. Because of the diversity of mental health team models, we now consider these separately (see the following section).

An additional two studies93,94 were identified in our updated scoping search (Table 9).

Levy et al. 93 describe a clinic in which a GPwSI ran an allergy clinic with a specialist allergy nurse. Ninety-one per cent of patients could be managed without onward referral and, even taking into account the referring GPs’ own estimates that one-third of patients referred to the community clinic would not have been referred to a hospital clinic, the authors still estimated a significant cost saving compared with referral to a consultant clinic in the hospital.

Moffatt et al. 94 describe a different model in which an endocrinologist visits practices on a regular basis, carrying out joint consultations with selected patients and their GP, and discussing cases with the GPs after the clinic session. Moffatt describes substantial educational benefits for the GPs, including self-reported reductions in referrals. Although no formal economic evaluations have been reported, these approaches appear costly and in general have depended on the enthusiasm of individual specialists to undertake this type of work. Their potential benefits in terms of longer-term improvements in quality of primary care and reduction in specialist referrals have not, to our knowledge, been evaluated. We describe other models of specialist liaison later under CMHTs, academic detailing and professional education. The potential for this type of arrangement to develop in future is uncertain, but there is interest in developing more formal arrangements, for example through the appointment of community geriatricians. It will be important to evaluate such arrangements in future in comparison with the appointment, for example, of hospital-based specialists.

Community mental health teams

In our previous review we commented on a 1998 Cochrane review comparing CMHTs with standard hospital-based care for patients with severe mental illness. This review found that CMHTs had reduced mortality rates, especially for suicide, fewer patients dropped out, there were reduced hospitalisations and CMHTs appeared to lead to reduce health service costs. 95 A broader review suggested that, compared with hospital-based clinics, CMHTs led to increased user satisfaction and improved adherence to treatment. 96

The NHS landscape has changed considerably since those reviews, with a multiplicity of models of mental health teams working in the community, mostly multidisciplinary, sometimes clearly specialist in orientation (e.g. for early-episode psychosis) and sometimes including more generic staff, including a category known as ‘primary care mental health worker’. These models operate in different ways in different localities and it is, therefore, difficult to describe and evaluate such services. The difficulty is compounded by the fact that, in many areas, an outpatient ‘referral’ has been replaced by referral to the local CMHT, when the patient may be seen by a range of possible professionals depending on the presenting problem. Bower and Gilbody97 described different models of specialist–primary care interaction and concluded that:

• Training of primary care staff was ineffective in improving outcomes unless intensive.

• Consultation-liaison could affect the behaviour of primary care clinicians but was difficult to generalise (note that consultation-liaison in this review refers to mental health specialists entering into an ongoing educational relationship with primary care clinicians. It differs from the use of the term in secondary care, which refers to psychiatrists working at the interface with acute general medicine, e.g. in the management of suicide attempts).

• Collaborative care (which had a number of definitions) led to small to medium-sized improvements in patient outcomes, patient satisfaction and patient compliance. These models generally involved drug treatment for patients with more severe disorders.

• Replacement/referral, including to a range of psychological therapies, were generally effective, at least in the short term.

• There was limited evidence on the cost-effectiveness of the models.

In the current review, we identified 10 papers,97–106 including four systematic reviews,98,99,101,102 which add to these findings (Table 10).

Christensen et al. 99 reviewed 55 controlled trials of treatments for depression. They concluded that a key predictor of successful treatment was a case manager providing direct feedback to the primary care physician, delivering psychological therapy and delivering an intervention that incorporated patient preferences. Nurse-, psychologist- and psychiatrist-delivered care were effective, but pharmacist-delivered care was not. Morgan et al. 105 confirmed that practice nurses could act as effective case managers for depression, this time in patients with comorbid diabetes or heart disease. However, Kendrick et al. 104 found that specialist mental health nurse support was no better than support from GPs for patients with ‘anxiety, depression and reactions to life difficulties’ in terms of patient outcomes, although patient satisfaction was greater in the mental health nurse group than in the GP group. A feature of this trial was that patients were enrolled with relatively minor symptoms (minimum score of 3 on the General Health Questionnaire; duration of symptoms at least 4 weeks but less than 6 months). It may be that specialist intervention is more effective for those with more severe symptoms.

Harkness and Bower102 conducted a systematic review of 42 studies on the evidence for the effectiveness of programmes in which on-site mental health workers, such as counsellors or psychiatrists, worked alongside physicians to provide therapy to patients. They found that primary care mental health works delivering psychological therapy and psychosocial interventions caused a significant reduction in consultations, prescribing and referrals to specialist care. However, they found that the changes were modest, not always consistent and did not generalise to the wider practice population. They commented that the economic significance of their work was unclear, but their results lent no support to the suggestion at the time that providing mental health care in primary care would lead to a substantial reduction in the cost of referrals.

A systematic review of 11 economic evaluations of enhanced primary treatment for depression found that these treatments, although effective, were associated with increased cost. 101 The authors estimated the benefit at between £7 and £13 per depression-free day.

A systematic review and meta-analysis of five evaluations of consultation-liaison psychiatry found no evidence of effectiveness of this intervention on antidepressant use or depression outcomes. 98 Although consultation-liaison models do have the potential to reduce referral, as in Younes et al. ,106 they are often dependent on the commitment of an individual psychiatrist, are hard to generalise and have not generally been subjected to economic analysis.

The siting of mental health workers and teams may be important in promoting collaboration, with Fickel et al. 100 commenting that collaborative care was often suboptimal, with the ‘collaboration’ occurring a most basic level. Kates et al. 103 found that the colocation of mental health workers in Canadian primary care practice improved collaboration.

We excluded a number of American papers from our review because the health-care system for mental health care is so different from that of the UK. We noted, however, the conclusion of one systematic review of US trials of integrated approaches to the management of depression involving primary care and mental health services. 107 They concluded that ‘Although most trials showed positive effects, the degree of integration was not significantly related to depression outcomes. Integrated care appears to improve depression management in primary care patients, but questions remain about its specific form and implementation’.

Telemedicine

There is a large literature on telemedicine, and in our previous review we included five systematic reviews and 20 studies. We concluded that few studies were designed in a way that allowed the overall impact of telemedicine interventions on health service use to be assessed. There were many anecdotal reports of outpatient visits being reduced by teleconsultations, and it seemed clear that a substantial proportion of teleconsultations do not need to be followed by a visit to the hospital for a face-to-face consultation. However, the cost-effectiveness of individual telemedicine interventions was highly context specific, and examples are available in the literature of instances in which telemedicine consultations were both substantially more expensive and substantially less expensive than conventional outpatient visits.

An additional 32 studies were identified in our updated scoping search,66,108–138 including three literature reviews108,116,136 (Table 11).

Many of the additional papers we identified had minimal evaluative components, but often showed the enthusiasm of the authors for a new development they had piloted. We have not reviewed each study in detail, because in large part they do not add greatly to the conclusions of the previous review and systematic reviews. However, we draw some conclusions from studies in the following areas.

Telecardiology involved ECGs being transmitted electronically and reported back to the GP by telephone by a cardiologist or nurse specialist. 137 With the increasing sophistication of automated analysis by the ECG machine itself, it is not clear that this is the ‘revolution’ claimed in the commentary by Backman et al. 108

With regard to teledermatology, the majority of evaluations were telemedicine for skin conditions, mainly of ‘store-and-forward’ services in which the GP sends a digital image of the skin lesion with a clinical description of the patient and the specialist responds with a diagnosis and management plan. In general, there is a reasonably high concordance between diagnoses made face to face and ones made from a good digital image, better concordance in terms of management plan and often sufficiently good concordance to be able to manage the patient with the potential to reduce referrals. 109,114,115,123,124,130,135 Telemedicine consultations may also reduce the need for outpatient attendance. 118,119

The clinical case for sending digital images for suspected skin cancer (teledermoscopy) is less clear, with two studies reporting that 49% and 70% of patients would require a face-to-face consultation110,127 and a high proportion of telemedicine diagnoses which were unclear or incorrect when compared with a face-to-face assessment. 125 A study by Tan et al. 133 reported discordance in diagnosis of 12.3% between teledermoscopy and face-to-face assessments for patients referred with possibly malignant skin lesions, with one malignant lesion missed by using teledermoscopy. Other authors reported safe and effective use of teledermatology for diagnosis of skin lesions,122,128 with the potential for significant reductions in the intervals between referral and diagnosis. 132 However, this result may not be relevant to the UK, where there is a maximum 2-week wait for outpatient appointments for suspected cancer.

Ophthalmology referrals were reduced in Scotland in a scheme which enabled optometrists to send a digital image along with a referring e-mail. Along with a redesign of the system pathway, a reduction of 37% in referrals was claimed, though it was not clear how much of this related to the pathway redesign and how much to the ability to send digital images. 111

In terms of teleotology, transmission of digital images of the eardrum along with audiometry and tympanometry enabled ENT specialists in Australia to give confident management advice. 117 This is perhaps most relevant to very rural locations, as audiometry and tympanometry are not currently widely available for children in UK primary care; they are provided in Australian primary care because of the very large distances that some patients would otherwise need to travel for the tests to be carried out.

For diabetes, Siriwardena et al. 139 identified 27 studies aimed at improving care. These mostly involved some form of behavioural therapy through videoconferencing or telephone calls. The sole outcome they reported was improvement in haemoglobin alpha 1 (HBA1), which they found improved in 12 studies (44%). However, the magnitude of change in HBA1 was not reported, and no economic evaluation was reported.

Telepsychiatry consultations have been widely reported with the aim of reducing travelling times in remote areas, but also to reduce the stigma associated with attending a psychiatric clinic, which may deter some patients from attending. Sometimes the primary care practitioner may be present at the consultation, an approach which is uncommon but may enhance the education value of the medium. Alternatively, the GP may present the patient at the start and discuss the specialist’s opinion at the end. 121 One RCT reported no differences in overall depression or in the number of days that were lost or unproductive owing to depression when telepsychiatry was compared with treatment as usual and for follow-up care. 113 A review of telepsychiatry for children and adolescents pointed to the weakness of the evidence base and could draw only the rather limited conclusion that ‘there are no data that suggest that this process contributes to negative outcomes’. 116

Live video consultations are reported in other disciplines, too, and are particularly likely to be favoured in very rural areas. 120 These studies appear to have been designed without a specific educational component, unlike the study reported by Thind et al. ,134 in which teleconsultations were established specifically to support a GPwSI in dermatology working in a remote area of Scotland.

Economic analyses are reported infrequently. A systematic review of economic analyses of real-time video communication concluded that teleconsultations, where the GP and specialist were both present, were likely to be more expensive than conventional outpatient clinics and not cost-effective for local delivery of services. 136 Our previous review included the only major UK study of ‘virtual outreach’, in which telemedicine consultations across a range of disciplines were found to be more expensive than conventional outpatient visits (£724 and £625 per patient, respectively). 140 In an economic analysis of a study in the Netherlands,118 the authors found that teledermatology became cost-effective when the patient was more than 75.1 km from a dermatologist, or when more than 37% of outpatient referrals could be prevented (which was judged to be unlikely). Likewise, van der Heijden found that 68% of dermatology referrals could be prevented by a telemedicine consultation (in which the GP was normally present) but costs were reduced by only 18%, and the authors were able to conclude only that the service could ‘probably’ be provided at lower cost. 135

Economic analysis of video consultations cannot be applied to the more common ‘store-and-forward’ approach of sending digital images to a dermatologist, which should, in principle, be cheaper. A number of authors, including Moreno-Ramirez et al. 129 and Butler and Yellowlees,112 have reported that patients can be managed more cheaply by ‘store-and-forward’ telemedicine than by conventional outpatient appointments. 112,121 However, the cost of establishing telemedicine services is not always included in published analyses. Furthermore, the logistics of setting up telemedicine services are significant and may make them difficult to maintain. 124 Sometimes the continuation of a service depends on the availability of enthusiasts. 138

One potential benefit of video consultations is the learning opportunity that they offer for the health professionals involved. However, although some studies report educational benefits (e.g. Hilty et al. 121), these are not always realised, and MacFarlane et al. 126 reported disappointing results in terms of GPs learning from the joint consultations.

The majority of applications conventionally associated with telemedicine require significant investment in fixed devices (e.g. video cameras). Quinn et al. 141 described the use of smartphones by district nurses to transmit images from the patient’s home. Although more a proof of concept that a formal evaluation, the authors suggest that smartphones have the potential to reduce clinic visits substantially. Two recent reviews of mobile phone technology in health care142,143 point to substantial opportunities for using smartphones in health care, well beyond the scope of this review.

Telecare

Telecare, defined as ‘offering remote care, often using sensing devices, of old and physically less able people, enabling them to remain living in their own homes’, did not form a major part of this review. As we commented in our previous review, ‘no wider conclusions can be drawn on the impact on outpatient attendance of telecare interventions, as this outcome was not generally reported’.

This remained the case because virtually all of the studies identified in our literature search focused on avoidance of hospital admission and did not report on outpatient attendance. For example, a recent major RCT of telecare in England (the ‘Whole System Demonstrator’) was focused on admission reduction, although they also reported that there was no difference in outpatient attendances between control and intervention groups. 144

We found that one study by Willems et al. 145 compared reorganisation of a hospital service to provide nurse monitoring of asthma through telecare with conventional outpatient appointments. They concluded that there was no significant difference in clinical outcome or secondary care usage between groups, although the type of professional accessed changed and there was no economic evaluation. We found no studies that specifically aimed to provide telecare as an alternative to outpatient attendance. Overall, we do not think it is appropriate to report on the impact of telecare interventions as an alternative to outpatient services, as that is not the target of the interventions, and outcomes relating to outpatients are commonly not reported.

Shared care

In our earlier review, we concluded that shared-care arrangements generally produced similar outcomes to conventional outpatient care but that evidence on cost-effectiveness was mixed, with some evaluations showing cost savings and others finding increased costs. We commented that successful operation of shared-care schemes depended on good communication between specialists and generalists.

Among the eight additional studies identified in this review (Table 12),98,146–152 we found one Cochrane review152 on the effectiveness of shared care in chronic disease management which identified 20 studies, including 19 RCTs. In this review Smith et al. identify five models of shared care:

• Basic model: a specific, regular communication system is set up between specialty and primary care. This may be enhanced by an administrator who organises appointments and follows up and recalls defaulters from care.

• Community clinics: specialists attend or run a clinic in a primary care setting with primary care personnel. Communication is informal and depends on the specialists and primary care team members meeting on site.

• Liaison: a liaison meeting attended by specialists and the primary care team in which the ongoing management of patients within the service is discussed and planned.

• Shared care record card: a more formal arrangement for information sharing where an agreed data set is entered onto a record card which is usually carried by the patient.

• Computer-assisted shared care and electronic mail: a data set is agreed on and collected in both the specialty and primary care setting and is circulated between the two sectors using computer systems such as a central repository or e-mail. This system may also include centrally co-ordinated computerised registration and recall of patients.

Smith et al. ’s review152 specifically excludes structured disease management programmes and specialist outreach clinics, which we also cover elsewhere. The comparator of ‘usual care’ was most frequently management delivered in primary care but in four studies the comparator was a secondary care clinic and we focus on these four studies in this review. They show no significant differences between shared care and outpatient care for clinical or psychosocial outcomes apart from recording risk factors which was better in the shared care model in two of the studies. For treatment satisfaction there were no differences, apart from one study of asthma care in which patients preferred hospital follow-up. Generally, the results support Smith et al. ’s overall conclusion from their review, which is that there were few identifiable differences between shared care and usual care. They interpret this as ‘no case for shared care’, but, in the context of the present review, ‘no difference’ between hospital follow-up and shared care with primary care is a positive result which shows the potential of the model. Costs are, therefore, key to interpretation of the findings; here, three out of the four studies in which outpatient care was the comparator show cost savings for the health service and one showed an increase in health service costs. In the three studies in which they were measured, cost savings for patients were substantial in the shared-care arms of the trials.

The majority of additional studies identified in this review provide a generally low level of evaluation, but in the main support the use of shared care in the management of long-term conditions. Only two of these present significant information on costs. The first of these presents the results of the economic evaluation of a RCT of intensive hospital outpatient management of patients with rheumatoid arthritis. 147 The mean cost per patient was slightly higher in the shared-care group, but a small gain in quality of life in that group means that the shared care intervention was likely to be cost-effective (at £2000 per QALY). In an earlier economic evaluation of the follow-up of people with serious mental illness,151 the costs of the shared-care model were critically dependent on the ‘amount’ of shared care offered, with complex patients receiving ‘medium’ or ‘high’ levels of shared care being significantly more expensive than those receiving ‘low’ levels of shared care.

We included consultation-liaison models in this review also, where there are typically joint consultations with specialist, patient and GP. All five studies identified were in mental health care, including the systematic review by Cape et al. 98 which we have already described in Community mental health teams and which found no evidence of effectiveness of the intervention on antidepressant use or depression outcomes. Given the significantly increased costs of this model, it does not present a cost-effective form of shared care, quite apart from the logistic difficulties of sustaining the model that we have commented on before.

Three qualitative studies looked at the problems which health-care teams experienced in developing shared-care models. The first related to a model of shared care in which a consultant wishes the GP to prescribe a specialist drug with which the GP may not be familiar. This type of shared-care model usually involves a protocol, for example detailing blood checks which need to be carried out. In this study, Crowe et al. 146 found that the GPs reported uncertainty and confusion surrounded the sharing of test results between primary and secondary care, which resulted in test duplication and omission. The GPs complained particularly about the lack of compliance of other staff in their practices and hospital colleagues with these arrangements (hospital specialists were not interviewed as part of the study). In a study in a very different setting, DeMiglio and Williams148 explored the problems in setting up shared-care arrangements for palliative care teams in Canada. Here, they found that working arrangements were inhibited by rules and regulations in the various organisations and by inflexibility in funding arrangements. In another Canadian study, numerous factors were found to inhibit shared care, including lack of resources, lack of time and incentives for collaboration and inappropriate GP payment models. 149

Guidelines, including referral pro formas

In this review we included two additional systematic reviews154,155 and 18 additional studies,75,156–172 shown in Table 13.

In their Cochrane review, Akbari et al. 154 concluded that passive dissemination of guidelines was ineffective, but that guidelines associated with structured referral sheets were associated with improved pre-referral management of patients. A second systematic review of the effect of guidelines on referral found that, in four RCTs, guidelines improved the appropriateness of pre-referral care, and in another trial included in the same systematic review there was an improvement in the appropriateness of referrals, although there was no measurable effect on practitioner knowledge, rates of referral or costs. 155 However, in none of these five studies was the guideline development ‘passive’; in each case it was combined with another intervention such as a structured referral sheet or an investigation pathway developed locally between GPs and specialists.

The limited effect of guidelines on their own is consistent with the wider literature on quality improvement. 173 Perhaps mindful of this, the great majority of the studies we identified combined guidelines with other interventions, for example referral pro formas,163,166 outreach or liaison visits by consultants,164,168,171 feedback on referrals,157 professional education,158 a more rapid appointment with the specialist if guidelines were followed159 and returning referrals which did not meet the guidelines. 161 All of these studies showed some impact of the intervention, although in some cases the improvement was confined to better information in the referral letter. Guidelines may also improve the quality of care provided without necessarily changing patterns of referral as, for example, in the urology shared-care protocol described by Spatafora et al. 171 The limited impact of guidelines on their own is not surprising, especially bearing in mind Clarke et al. ’s conclusion from an empirical survey of GPs that referral guidelines were ‘rarely used in practice’. 155

However, there were other studies which did not report any change, for example incorporation of guidelines into a pro formas for head and neck cancer165 or for mental health problems,170 and three studies of passive dissemination of referral guidelines showed them to be ineffective. 167,169,172 Slade et al. 170 commented on a number of reasons why the pro formas in their study may have been ineffective. These included GPs feeling that the pro formas were oversimplistic, did not reflect the complexity of patients’ problems and were seen as a means of reducing referrals.

Guidelines can also be used to reject referrals. Garcia Garcia et al. 160 reported a study in Spain in which nephrologists rejected referrals which did not adhere to guidelines, sending 28.5% of referrals back with ‘an explanatory report including suggestions for patient management’. They found that the GPs included in the programme ‘showed great interest and no complaints were registered’.

Although some guidelines have been developed to reduce the number of inappropriate referrals, it should not be assumed that the aim or outcome is always a reduction in secondary care utilisation. Hamilton et al. 162 provided practices with risk assessment tools to assist GPs in selecting patients for cancer investigation and found that their use resulted in an increase in the number of chest radiographs ordered, the number of referrals and the number of cancers diagnosed.

Bearing in mind the well-established literature on the relative ineffectiveness of passive use of guidelines in changing clinical practice, guidelines should in general be regarded as part of other interventions. Indeed, they are an essential part, and it would not be realistic to conceive of many of the interventions considered in this section unless they were underpinned by clinical or organisational guidelines. The place of referral pro formas as a way of reinforcing guidelines has increased rapidly in recent years as electronic record systems have become more sophisticated. These pro formas are sometimes used as a way of requiring information from the GP before a referral request can be accepted, and this may include the results of tests that locally agreed guidelines indicate should be done before referral for a particular condition. An example might be that the result of a partner’s sperm count should be available when referring a woman with infertility. Depending on whether information in a referral pro forma is advisory or mandatory, pro formas can be used to require adherence to guidelines. An example of a guideline being used to incentivise particular professional behaviours was reported by Fleuren et al. :159 if Dutch GPs adhered to an agreed protocol for the first 6 weeks of back pain, a priority referral and MRI were guaranteed if the patient was subsequently referred. This led to a reduction in referrals judged as inappropriate against national guidelines for referral from 15% to 8%. 159

Audit and feedback

We found few relevant papers157 on the impact of audit and feedback on referral behaviour. In our previous review, we concluded that simple feedback of rates of referral had no impact on referral practice.

In the present review, no studies of simple audit and feedback were included in the Akbari et al. Cochrane review. 154 Elwyn et al. 157 reported a study in which feedback was sent to GPs on the appropriateness of their referrals to endoscopy against NICE referral guidance (but with no referrals being denied). The intervention was associated with an increase in the appropriateness of future referrals.

Despite the limited support from research findings, feedback of rates of referral without any indication of appropriateness remains common, and probably ubiquitous, across PCTs and CCGs in England. If such feedback is to have any effect, it probably relies on GPs having the time and the inclination to review the clinical appropriateness of their referrals.

Professional education, including academic detailing

In this section we include all forms of professional education, including ‘academic detailing’, which is a term used to define a particular type of face-to-face engagement between an expert (e.g. a specialist) who may visit a practice on one or more occasions. It has most commonly been used in relation to prescribing, with the term ‘academic’ principally indicating that there is no pharmaceutical company involvement.

We identified 18 studies154,164,174–189 which assessed the effect of professional education schemes (Table 14).

Akbari et al. 154 included three RCTs of specialist-led educational strategies, one of which was highly intensive, involving monthly visits by orthopaedic surgeons to discuss cases in which the GPs were uncertain of diagnosis or management and were considering referral (the others were less intensive). The most intensive of the interventions was associated with a near halving in the number of orthopaedic referrals. Of the other two, both interventions were associated with higher rates of referral following the educational intervention (though the reviewers note a unit of analysis error in one of the studies which might have led to an overestimate of statistical significance). Both of these results were against the hypothesised direction of effect.

Evans et al. 178 described a scheme in north Wales in which GPs had weekly protected time to discuss the appropriateness of their referrals, with regular input from local consultants. For orthopaedics the scheme was associated with a reduction in referrals by up to 50% and an increase in use of other local services, with a better understanding of other local management options for common orthopaedic conditions. An observational study by Cameron et al. 176 in the USA evaluated academic detailing as a strategy to increase the early detection of dementia in primary care practices. In a follow-up evaluation, 55% of the doctors said that they had increased referrals to community resources, 24% said that they were already making satisfactory referrals prior to the visit and 21% said that they had not increased referrals but intended to do so.

Suris et al. 184 described an intensive biweekly programme of support by rheumatology specialists which was associated with a 31% reduction in referrals. Other studies also reported a reduction in number or an increase in the appropriateness of referrals following educational programmes. 175,177,180,182,186,188,189 However, some studies found little or no impact of education on the rates or appropriateness of referrals. 164,174,181,183 It should be noted that some of these involved a relatively modest intervention, for example a lecture or a single outreach visit, and in one case GPs were simply sent a 12-page illustrated guide to common skin lesions. 183 A study by Ulrik et al. 185 demonstrated that the focused education of GPs and their staff improved the diagnosis and management of COPD, but did not significantly alter the percentage of referrals. It is probably important to distinguish programmes whose main aim is to improve quality of care without any specific aim to reduce (or increase) referrals.

Most of the interventions described here were aimed at increasing the appropriateness of referrals or reducing the number of referrals, or both. However, we note that in two of the three trials included in the Cochrane review, referrals were increased following the educational intervention. It is important to acknowledge that better clinical management in primary care will sometimes result in an increase in specialist referral. We noted some time ago that GPs with greater knowledge in a clinical area tended to refer more patients to hospital in that area,190 and this possibility should not be ignored at a time when local pressures are likely to be exclusively focused on reducing referrals.

In their review, Akbari et al. 154 described professional education as ‘potentially effective’, and this is probably a reasonable summary of a literature in which studies and interventions are very varied, ranging from intensive specialist support to single educational events and, although we have not analysed this in detail, the more intensive or multifaceted interventions appear to be more effective. 154 There is a clear tension between education at an intensity which is unlikely to be sustained (e.g. monthly practice visits from specialists) and could not be mounted across a range of specialist disciplines, and more modest interventions that appear less effective. There is a question of whether or not the impact of an intensive intervention can be sustained over time, thus allowing rotation of specialists into an ongoing set of educational programmes. We found no evidence on the long-term sustainability of changes associated with educational interventions.

In-house review of referrals

In our previous review, we commented that ‘newer innovations, such as obtaining an “in-house” second opinion prior to referral, appear promising but require further investigation’. Akbari et al. 154 drew the same conclusion, but still based on only one study. 191,192 In our current review, we identified four additional studies that examine in-house review of referrals (Table 15). 178,179,193,194

We identified four additional studies in our review. Two papers by Evans178 and Evans et al. 179 reported on an intervention in which GPs discussed the appropriateness of their referrals on a weekly basis, observing a reduction in the variability of rates of referral between doctors in the practice, and a reduction in the overall mean rate of referral. The approach also appeared to be acceptable to patients. The arrangement described by Evans et al. is not the same as either that in the previous study, in which referrals were routinely seen by another GP before the decision to refer was confirmed, or the model which has been taken up and incentivised by a number of PCTs/CCGs, in which referral letters are scrutinised by other GPs in the practice before being sent to the hospital.

In a report on referral management published by The King’s Fund in 2010,194 including a literature review and case studies, the authors concluded, based on limited evidence, that, of the referral management strategies considered, a strategy ‘built around peer review and audit, supported by consultant feedback, with clear referral criteria and evidence-based guidelines is most likely to be both cost- and clinically-effective’. However, more recently, a retrospective time-series analysis of outpatient attendances from practices which started implementing internal peer review of referrals showed no significant changes in outpatient attendance rates. 193

Because of the interest in these models and the lack of published literature, we conducted a substudy of in-house referrals as part of this project, and this is reported elsewhere in this report (see Chapter 4).

Referral management centres

In-house referral schemes discussed in the previous section involve some form of scrutiny of referrals within the referring GP’s own practice. RMCs involve the scrutiny of referrals outside the practice. They have been introduced in an attempt to ensure that referrals are appropriately directed and (though not always explicitly) to produce downward pressure on GPs’ rates of referral. These organisational interventions are new and they are not mentioned either in our previous review or in the systematic review by Akbari et al. 154

In our current review, we identified six studies193–198 about RMCs (Table 16).

Referral management centres were first introduced for orthopaedic problems in order to deal with the large number of referrals to orthopaedic surgeons which they felt could be better managed by other practitioners (e.g. physiotherapists). GPs did not refer to these other services either because they were unable to or because they were unaware of them, and so some RMCs were established with the extra capacity to provide additional care in the community (i.e. they did more than just triage referrals). For example, the Bath and North East Somerset Orthopaedic Interface Service retained 50% of referrals passed to their system, sending only 21% on to see an orthopaedic surgeon. 196

However, other RMCs have been established to screen all referrals, with an explicit or implicit aim of reducing hospital outpatient attendance. The aim of these services is to identify inappropriate referrals, which can be considered in terms of three broad categories:

• referrals made to the wrong service or specialist

• referrals containing insufficient information, making it difficult to assess urgency or relevance

• referrals that do not conform to accepted clinical guidance.

Referrals may be ‘accepted’ by the RMC and passed on to the hospital clinic, diverted to a service judged more appropriate or returned to the GP as inappropriate. Considerable concerns have been expressed about these centres in terms of introducing delays in treatment, interfering with the GP’s clinical judgement, making inappropriate triaging decisions based on information in the letter without seeing the patient, breaching patient confidentiality and restricting patient choice. 195

Although we found studies describing concerns about these centres,194,198 we found only one published article in the academic literature of an evaluation of RMCs. This involved three centres which were set up to reduce referrals from GP practices in Norfolk. 193 The introduction of two of these had no impact on numbers of outpatient referrals, whereas the third was associated with an increase in referrals. Comparison with two schemes involving within-practice peer review of referrals indicated that RMCs were more expensive but not more effective than these schemes.

Although not strictly relevant to the UK context, one American study was also noted, in which all referrals from primary care were directed through an internet-based ‘branching logic that required a series of clinical responses regarding the patient’s diagnosis or symptoms’. This resulted in the rejection of 22% of referrals as inappropriate. 197

In addition to these peer-reviewed articles, two further reports on RMCs were identified in the grey literature. The first, a report on an evaluation of seven referral management pilots in Wales,198 reported that although there was significant variation in the quality and potential sustainability of the initiatives, which were diverse in their scope and operational details, the majority of pilots provided some ‘hard’ evidence of the impact of their referral management diversions. Pilots that were designed to tackle single issues or groups of related problems were considered to be more successful than those using an all-specialty generic model of referral management, which was not considered to be cost-effective. The second report, on a study conducted for The King’s Fund,194 examined approaches to referral management, including RMCs, more broadly. This study included a literature review, two phases of qualitative research (interviews with 21 commissioners and case studies of four schemes, including two RMCs) and quantitative analysis of outpatients attendances. The authors that concluded that RMCs were less likely to represent value for money than more passive approaches such as peer review.

Because of the interest in these models and the lack of published literature, we conducted a substudy of RMCs as part of this project, and this is described elsewhere in this report (see Chapter 3).

Financial incentives

Interest in financial incentives in relation to referrals falls into two areas:

• direct financial incentives for GPs to reduce referrals

• secondary effects on referrals of other financial incentives.

We do not consider here the effect of fundamental changes in remuneration systems for doctors in primary care, for example changing from fee for service to capitation.

We included six studies in our review (Table 17). 154,199–203

The Cochrane review by Akbari et al. 154 described two studies of the impact of GP fundholding on referrals in the 1990s. In this scheme, GPs had a direct financial incentive to reduce referrals for conditions for which they had a budget (e.g. elective surgery). There appeared to be a small reduction in rates of referral in fundholding practices. In our previous review, we also suggested that financial incentives introduced for GPs to conduct minor surgery in the early 1990s led to them carrying out operations for relatively trivial conditions that would not previously have been referred to hospital or where non-operative treatments were available, a conclusion supported by a more recent study. 202

A second set of studies assessed the impact of the QOF on hospital utilisation. 199–201,203 This pay-for-performance scheme did not (initially) have any incentives related to outpatient referrals but did provide incentives for chronic disease management that could have impacted on referrals. The evidence is clear that it did, with a step increase in referrals to renal clinics after the introduction of an incentive to identify and manage patients with renal failure,199,201 and an increase in referrals to a diabetic clinic when tighter targets for blood sugar control were introduced into the QOF. 203 Referrals also increased to neurology clinics following an incentive to provide better seizure control for patients with epilepsy,200 although in this study a neurologist provided additional educational support to practices at the same time.

More recently, an incentive has been introduced into the QOF for GPs to:

• review data on secondary care outpatient referrals

• participate in an external peer review with other GPs in the CCG to compare its secondary care outpatient referral data with those of the other contractors

• engage with the development of and follow care pathways to improve the management of patients in primary care in order to avoid inappropriate outpatient referrals.

No data are available on the impact of this incentive on patterns of referral and there are currently no incentives to actually reduce rates of hospital referral.

Specialist advice as an alternative to referral

The traditional model of outpatient referral in the NHS is for the patient’s GP to write a referral letter, after which the patient is given an appointment and seen in the outpatient clinic. In this section we review ways in which GPs can seek advice about the management of individual patients without the patient actually being seen by the person giving that advice. These developments are relatively new and were not covered in our previous review or in the systematic review by Akbari et al. 154

A number of studies have reported interventions in which the GP seeks advice by letter or e-mail but without an expectation of the patient being seen by the person giving that advice (Table 18). 204–211

Patterson et al. 205 described a system by which GPs could e-mail neurologists for advice, which they found safe and effective in reducing face-to-face outpatient appointments. There was some increase in overall referrals (i.e. e-mail and appointment requests combined) but an overall reduction in cost, with small increases in GPs’ time outweighed by reductions in specialist time. Similarly, Williams et al. 211 assessed an internet neurology referral system (Neurolink) for GPs in Ireland. The authors reported that 19% (127/662) of patients did not require a neurology outpatient appointment and, instead, the GP was given advice, treatment and investigation options for those patients.

In endocrinology, Oliva et al. 204 reported that 88% of queries from GPs could be resolved without the need for a physical consultation, and Walker et al. 209 also reported the success of an e-mail advice service for diabetes and endocrinology.

Stoves et al. 207 described a system by which the GP requested advice by e-mail on patients with mild to moderate chronic kidney disease but in a situation where the specialist had access to the GP’s electronic medical record. Ninety-two per cent of requests for advice could be managed without the specialist needing to see the patient, and, when the GP did request a face-to-face consultation, the majority could still be managed in primary care with the provision of appropriate advice.

The telephone may also be an underused resource. In an Australian example, psychiatrists saw patients only if it was judged necessary after a telephone conversation between the psychiatrist and the GP,208 and in the USA a telephone conversation between the primary care physician and a paediatric subspecialist meant that referral was avoided in 32% of cases. 210

There is overlap between the studies of electronic communication and some of the studies we describe in Relocation, Telemedicine, especially the use of store-and-forward images for skin problems and diabetic retinopathy. Sometimes a teleconsultation has been used to give advice to GPs. For example, Scalvini et al. 206 describe how a 10-minute teleconsultation with a specialist was able to resolve half of the problems in a group of hypertensive patients at high risk of cardiovascular disease.

Seeking advice from specialists by letter or by telephone is not a new approach. Indeed, over 20 years ago Roland and Bewley212 described an advice line (‘Boneline’) through which GPs could seek advice from orthopaedic surgeons by telephone. However, a number of things have changed which make it likely that telephone and electronic communication between GPs and specialists will play a much larger role in future. Telephone is perhaps the least easy method of communication to use, simply because it requires the availability of both the specialist and the GP at the same time. E-mail, however, provides the opportunity for asynchronous communication and can also be used to transmit test results, images, etc. Doctors, along with others in society, are used to getting rapid responses to queries using e-mail in their personal life. There is no reason why medicine should be different from other sectors in its use of this technology, and we note that a number of NHS trusts are already offering an electronic advice service with an agreed tariff with local commissioners.

There are a number of potential consequences of the ability to communicate more easily with specialists. One is that the need for face-to-face referrals could diminish, and several of the studies quoted above suggest that this could happen. However, easy access to specialist opinion could also increase the number of patients eventually seen in clinic. The potential benefits of this for patient care (i.e. seeking specialist advice which would not otherwise have been sought) are unknown, and this would be an important area for future research.

Decision aids

Patient decision aids were not considered in our previous review or in the systematic review on interventions to improve outpatient referrals. 154 In this review we identified one relevant RCT in which patients with a range of chronic conditions were offered coaching on how to discuss their condition with their primary care physician. Patients in the coaching group were much more likely to be referred to a specialist in subsequent consultations with their primary care physician than those who had not received coaching (51% vs. 28%). 213

Although the results of this study in the USA may not be directly applicable to the UK, they raise an important question of what the ‘correct’ level of referral is. The studies described in earlier sections of this review use professional benchmarks for assessing the appropriateness of referrals (e.g. NICE guidance). Even where professional opinion is the gold standard we noted some interventions in which improvements in care were associated with increases in hospital referral.

This study raises the additional issue of what patients would prefer, and it was beyond the scope of this review to identify whether patients think they are referred to specialists too often or not often enough. In making referral decisions, GPs assess the clinical need for referral, but they are also influenced by the patient’s wishes. 214 They are, therefore, balancing clinical need with the wishes of the patient and the availability of health-care resources.

Referral management centres had a wide range of aims and programme theories

The RMCs included in this study were selected on the basis that they aimed to reduce referrals. In all cases, however, this was just one of many inter-related aims articulated (summarised in Figure 2). At the highest level, interview data indicated that the broad aim of the schemes was to ensure the best use of limited resources, under which three chains of more specific subaims were described, highlighting the ways in which this could be achieved: quality improvement, reduction in referrals to secondary care and increased efficiency in the referral process.

FIGURE 2.

The aims and functions of RMCs, based on interview findings.

With respect to quality improvement, interviewees described how RMCs aimed to collect and analyse data on referral quality and, based on the understanding gained, to standardise referral processes, educate GPs and implement primary care pathways. Desired outcomes included improvement in the quality (and reduction in the variability) of referrals and, in turn, improvements in the quality of patient care, with patients seen ‘in the right place, at the right time, and by the right clinician or specialty’ (B4, GP with managerial responsibilities), ensuring that ‘when they’re sitting in front of that person, that person has everything that it’s reasonable and possible for them to have in order for them to treat that patient appropriately’ (B1, manager).

In order to change patterns of referral, interviewees described how RMCs aimed to collect rich data on referral patterns, to inform the development of alternative, more cost-effective models of service delivery to meet identified local needs, to divert referrals to alternative services (such as lower tariff community services) and, in so doing, to reduce referrals to secondary care. Also described was the expectation that improving referral quality would also result in fewer referrals, with the aim of educating GPs to provide ‘better quality letters where more information was there, and therefore justifying more effectively, the reason for the referral’ leading to GPs ‘doing more in the general practice setting’ (D3, GP triager).

Finally, to make the referral process more parsimonious, RMCs aimed to centralise referral and appointment booking processes, to hold comprehensive and up-to-date knowledge on available provider services, to reduce the (administrative and knowledge seeking) burden on referring GPs and, through these measures, to increase the efficiency of the referral process.

Interviewees described a complex network of aims and underlying tacit assumptions regarding how the achievement of one aim depends on the delivery of another.

[The aim was] to improve the quality of referrals, to reduce unwanted variation and to trust that by doing those two things an unplanned consequence would be [. . .] a reduction in unnecessary expenditure.

D1, manager, RMC provider

Yeah, certainly my thrust has always been about improving the quality [of referrals] [. . .] They can’t try and address the referral problem and not raise the quality of primary care.

A2, manager, commissioning, CCG

The weight given to the different aims varied significantly, not only between schemes but also between stakeholders within the schemes. While one GP user of a scheme (B6) perceived the aim to be to ‘save money on referrals into secondary care’, a manager involved in its commissioning (B1) was clear that it aimed ‘to collect data on why patients are referred’ to inform how best to ‘shape services for the future’. Similarly, a manager for a RMC provider organisation described a mismatch between commissioner and provider expectations.

[The CCG] saw the RMC as a way of reducing referrals, therefore reducing costs. So I had to explain to them ‘you’re not going to reduce costs by just commissioning a RMC in isolation’, for starters. [. . .] you will only reduce costs by commissioning community pathways at a reduced tariff. [. . .] I have to be very clear when commissioners come to us about what their expectations are of a referral management centre [. . .].

C1, manager, RMC provider

Referral management centres evolved substantially over time

Not only did the aims of the schemes (or the weighting of them) vary between sites and between stakeholders, but interviewees also reported that they had evolved over time, dependent on changing local priorities and on the successes already achieved.

So the idea has changed as the service changed. They were initially collecting that data [on referral quality] and being able to present that [to practices] and want to possibly performance manage GPs. [. . .] but more recently it’s been about making sure that patients, are seen in the right place, so not going into inappropriate clinics. Not going to clinics with investigations that could have been done in primary care, having to be done again in secondary care. It’s about improving the patient experience, that’s the intention.

B3, GP with managerial responsibilities

Interviewees described how the functions of RMCs had evolved considerably since their initial implementation, in response to changes in aims and local context. For example, for two of the schemes (RMCs B and C), data collected were reported to have informed the development of new community services, which in turn led to a need for the introduction of clinical triage (to divert referrals appropriately).

. . . first of all [. . .] we’d look at which specialties we thought would benefit from a community pathway, then we had to bring in the triage so it could be agreed which patients went into that community pathway. [. . .] No changes have been made because the Gateway wanted change if you see what I mean, they were all changes that were going on in the landscape with the CCG and the GPs, the providers, and then the Gateway responds.

C1, manager, RMC provider

Also highlighted was the fact that evolution in the functions and processes of RMCs resulted from efforts made to overcome identified challenges, for example to respond to GP concerns or iron out administrative glitches. For RMC A, for example, the bookings management element of the scheme was originally provided by an administrative team some distance away from the CCGs served, but, for issues of cost and the acknowledged benefits and increased acceptability of using staff with local knowledge, this was relocated in the CCG area.

The capacity to evolve was described by several participants as key to the success of RMCs.

[. . .] if it’s going to be a success, it needs to evolve, [. . .] you can start off with something but if it stays the same, then I think in the case of ours, we’d be having problems with it; whereas as it’s evolved it’s been able to respond to the needs of firstly the PCT and now the CCG.

B2, GP with managerial responsibilities

One interviewee acknowledged that, by encouraging appropriate referral behaviour on the part of GPs, RMCs might find that they were no longer needed.

[. . .] once the provider landscape has settled down, and people are much clearer about what they should be managing, and what they shouldn’t be managing, and what they should be sending, and what they shouldn’t be sending. If all that were in place, we might not need an RMS [. . .] it might not be something that we’ll have forever.

B1, manager, CCG

Software and system limitations

System design and information technology (IT) compatibility issues were reported to present a major challenge. Both a lack of clarity in aims and changes in the primary function of RMCs (e.g. from administrative to clinical function) presented IT challenges, with adaptations of software for new purposes causing difficulties for data capture and issues with the subsequent quality (and hence the utility) of the data collected.

I think that the problem was that the original set-off of the RMS, being admin managed, had a database that was designed to manage from the admin point of view. [. . .] So the clinical steps that were taken have been put on top of the old admin database. [. . .] the impact that that made meant that data protection was a bit harder, but also that audit of clinical information was harder.

B5, GP with managerial responsibilities, and GP triager

[. . .] it was an aspiration really, saying ‘we will use this data for commissioning intentions, commissioning purposes’. However, on looking at it, we’ve seen a few issues in the data quality, so we’re not prepared to use that as a reliance really. It’s good at giving a ball-park at the moment but we’ve tasked our data teams [at the CCG] to start, well, continue looking at it with them and get them up to speed as to what exactly we’re wanting to get out of it.

C2, manager, CCG

Some features of clinical and administrative IT systems were highlighted as being important enablers of the schemes and, conversely, their absence was highlighted as disadvantageous. For RMC B, the lack of a facility for GPs to carry out triage remotely was cited as a reason for difficulties experienced in recruiting local GP triagers, which in turn was reported to reduce the credibility of the scheme, thereby affecting GP buy-in.

Contract and capacity issues

A number of interviewees described difficulties with successful contracting of RMC services and managing capacity within RMCs, in the context of continual and unpredictable changes in service structure and demand, with the potential to have a negative impact on both the sustainability of the scheme and the quality of patient care.

[. . .] more triage specialties were brought in [. . .] That did result in a number of changes to process [. . .] The team kept saying ‘you are adding extra steps in, you are increasing the risk of delays’, because it took longer to invest the time to do each referral. [. . .] But that was the way the commissioners wanted it. So we were taking longer to process the referrals, the [RMC] team doubled in size [. . .], which meant a lot of new members joined it and then it’s a 5-week training programme, so that hindered progress. And the activity levels started to rise too. So the combination of all those factors made it very difficult for the team to keep up with the volume that was coming in and a backlog started to generate.

C4, practice manager also involved with RMC

Establishing shared vision

Both commissioners and providers of RMCs described a lack of clarity in the aims at the outset, with one interviewee commenting that those developing the RMC had been ‘shooting in the dark’ (C1, manager, RMC provider). Also highlighted was the challenge of managing the differing priorities of commissioners and referring GPs, and communicating effectively to sell the concept to all involved.

We had to speak a little bit with fork tongue [. . .] They [the CCG] were most interested in reducing the referrals but we said to them that, ‘if we speak to our constituency [GPs] about that subject we will get yawns and non-participation’.

D1, manager, RMC provider

Unclear and poorly communicated aims were associated with a perceived lack of awareness among referring GPs regarding the purpose of the intervention and of their role in it.

If I was going to do this again, there would be a huge advertising campaign [targeting referring GPs], so to speak, actually saying what the intentions are [. . .] I don’t really think that many of my colleagues really had a grasp of that at the beginning.

B5, GP with managerial responsibilities, employed by RMC as triager

As time’s gone on I think I’ve understood the aims, but it hasn’t been clearly set out by the RMS. I think I’ve learnt through the process what the aims are, but that’s my idea of what it is. I haven’t ever received anything from the RMS to say this is what we’re trying to do.

B6, referring GP

A lack of early and effective involvement of GPs in the development of the aims of RMCs and a perception of being ‘told what to do’ (B6, GP) was reported to contribute to misinformation and distrust and, consequently, resistance to engagement with RMC processes.

[. . .] if I had a conversation with doctors about actually what it is that we’re trying to do, then I often find that people are much more ready to come on board, more aware of the process. And that one-to-one dialogue, it would be much better if that dialogue had been done at an earlier phase. When you have misinformation out there, then it pollutes what people actually think and turns it into a purely financial-driven motive and actually the points about improving patient care, reducing waiting times and saving money at the same time will get missed.

B5, GP with managerial responsibilities, employed by RMC as triager

Sustaining engagement

Participants described a wide range of measures implemented to keep GPs on board (Table 20).

The evolving nature of RMCs, however, was reported to present significant challenges, with rapid and widespread changes occurring simultaneously in many different referral pathways, leading to frustration among referring GPs unable to keep up.

[. . .] the rules changed. I think, initially, it was some clinics, then it was other clinics and then it was everything and it was keeping up with their change [. . .] As they were developing, we were trying to understand what they were doing.

B6, referring GP

Clarity and parsimony of referral processes were described as important factors influencing GP support.

[. . .] the advantage is that now we don’t do Choose and Book it is a lot more straightforward at that interface with the patient; [. . .], so we saved a lot of time.

A4, referring GP

[. . .] as a GP I don’t want to be thinking about a lot of different pathways. Ideally I want to try and keep it down to as little as possible, so you know the more the RMC does, the better, because you just, you end up sending 90–95% of your referrals to the RMC and you don’t have to think about it after that.

C3, GP with managerial responsibility

Maintaining quality and safety in patient care was considered paramount to ensuring continued engagement. The potential for introducing delay into the referral process was acknowledged as a serious threat to the acceptability of the schemes.

We knew that the whole systems would start falling apart if clinical safety and quality fell apart [. . .] When GPs heard about it [breaches in agreed turnaround times for referrals] . . . I mean it wasn’t kept secret . . . then it was that loss of faith in something [. . .].

B2, GP with managerial responsibilities

Respecting general practitioner autonomy and responsibility

A major stumbling block in selling the concept of RMCs to GPs was the perception that such schemes present a challenge to clinical autonomy, the avoidance of which was cited by interviewees across all the included schemes as an important consideration in the design, implementation and sustainability of the initiatives (as outlined in Table 20). Even with respect to the RMC, however, this took a purely educational approach, with GPs retaining responsibility for the referral destination; interviewees described the challenges associated with GPs’ resistance to negative feedback.

[T]he tone in which the triagers rattled their notes into the computer, when they were finding a referral not as good as it might be, was not always as diplomatic as it could be. [. . .] we had to put that text through a moderator who would change all the language so that there were no [. . .] unnecessary spikes in it, [. . .] We train them [GP triagers] all in how to do that.

D1, manager, RMC provider

[T]here is rivalry in primary care. We all like to feel our practices are the best practice [. . .]. I know that a number of practices were being contacted fairly frequently [. . .] they hadn’t done this, that or the other; and I think that some practices didn’t like that and left, which were the very practices one could say that really required that support.

D2, GP with managerial responsibilities

Different measures of success for different stages of referral management centre development

Perceived success achieved in relation to the aims set out in Figure 2 could be seen to reflect, to some degree, the maturity of the scheme, as, for example, effective data collection and analysis was a prerequisite for the implementation of processes to achieve the specified intermediate and desired outcomes and move towards the overarching aim to ensure best use of NHS resources. Interviewees tended to report greater difficulties in measuring achievements against (and were less confident in their claims with respect to) higher-level goals, with success claimed less consistently, reflecting marked disagreements between stakeholders (as illustrated in Table 21).

For all schemes, RMC providers and commissioners expressed with varying degrees of confidence that referrals to secondary care were being curbed by the schemes. The calculation of efficiency savings, however, was reported to be a highly complex process, requiring sophisticated analysis, the factoring in of costs of running the RMC, the provision of alternative services to which referrals were diverted and possible disinvestments in secondary care. Providers and commissioners described the challenges of both calculating savings and communicating the findings, which were, thus, open to interpretation.

[. . .] we’ve got a schedule [for calculating savings] [. . .], in the absence of anything more specific, we’re trying to quantify it that way. Now whether there’s a better or more sophisticated way of doing it, I don’t know [. . .] I think we’ll always be looking at our data, and how it’s presented, and whether there’s a more accurate way of looking at things. But, I think on the whole it’s convincing people that it’s a good thing, and it’s the right thing.

B1, manager CCG

The trouble with differing priorities

Effectively communicating success and a shared vision of what this looks like for a particular scheme was held by many to be essential for the scheme’s survival. The provider of one scheme recounted how its failure to meet the commissioner’s aim to centralise and standardise referral processes (owing to a number of practices opting out), combined with a lack of confidence in the figures demonstrating cost-effectiveness, meant that despite strong evidence of success (from the perspective of the providers) the contract for the scheme was not renewed.

[The CCG] took the decision [. . .] that it was more important to bring all the practices together under a common denominator, even if it was the lowest common denominator, rather than to [. . .] look at the evidence and purchase that which is known to be effective. We were bitterly, bitterly disappointed . . . [. . .] we were very well resourced, we fulfilled all our expectations, we got going, we saw the curve beginning to bend [. . .] The big, big problem was that we as leaders didn’t operate in such a way that others felt drawn in [. . .].

D1, manager, RMC provider

Evidence of effect

There is some evidence from the countries reviewed that the use of financial mechanisms has impacted referral rates and the dependence on hospital-based service delivery more broadly. However, much of the evidence pertains to evaluations of local or pilot initiatives, and there are few robust data on their long-term impacts at the system level. For example, there is evidence from Denmark that following the change by Copenhagen municipality from a mostly capitation-based system to one that combines capitation and fees in the 1980s, there was an increase in the volume of activities that were specifically remunerated and referrals to specialists decreased,276 although it is unclear if this reduction has been observed elsewhere across the country and whether or not it was sustained long term.

In the Netherlands, a small number of evaluations of interventions are linked to fee payments under ‘modernisation and innovation procedures’. For example, one observational study of 70 GP practices of the substitution effect of transferring minor surgery to primary care found evidence of a significantly lower referral rate for minor surgery for patients with sebaceous cysts over the period 2006–10, although the overall effect was small (odds ratio 0.98, 95% confidence interval 0.97 to 0.99). 277 An evaluation of a teledermatology service involving over 37,000 teleconsultations conducted between 1820 GPs and 166 dermatologists over a period of 3.5 years found that 68% of referrals to a dermatologist could be prevented (see also Chapter 2, Findings, Relocation). 135 The estimated cost reduction was 18%. However, related work also suggested that GPs were facing challenges in identifying suitable patients for this approach. 270 Available evidence on teleconsultation in eye care and cardiology in the Netherlands has pointed to reductions of referral rates of around 50% for patient selected by the GP for telemedicine;270 similar observations were made for telepulmonology consultations. 278 However, in Chapter 2 (see Findings, Relocation, Telemedicine), there is a general lack of economic evaluations which take a whole-system perspective.

The evidence on the impact of bundled payment for selected chronic conditions through mostly GP-led care groups remains generally mixed. Care groups in the Netherlands subcontract GPs, medical specialists, nurses and other disciplines, and approximately 78% of GPs are members of such a group. 279 Recent evaluations indicate that the system of bundled payments improves the organisation and co-ordination of care. 280 However, thus far there is little robust evidence on the impacts of care groups on referral patterns. 281

In Spain, there is very little robust evaluation of the impacts of financial mechanisms and incentives on secondary care referral. Primary care providers have argued that the effectiveness of such incentives has decreased because of successive reduction of the incentives. 282

For Finland, the 1993 state subsidy reform formed part of a larger move to create economic incentives for municipalities to improve the efficiency of services. 257 A number of changes have occurred in the organisation of primary care as a consequence of these reforms but there are few robust data that would allow for the assessment of its impacts on secondary care referrals.

Evidence of effect

In Spain, measures such as transferring minor surgery to primary care settings and offering direct access to tests and investigations are generally seen to be effective approaches in reducing the referral of patients to specialists. 284 However, anecdotal evidence suggests that, although widely available, minor surgery procedures in primary care do not seem to have reduced pressure on hospital surgery departments but instead might have led to an increase in service utilisation, by uncovering a previously unmet need for minor ailments (such as warts or verrucae).

The use of practice assistants or nurse-led diabetes treatment in the Netherlands has been linked to a reduction of 40% in referrals to hospital. 269 However, this transfer has not led to a reduction in total diabetes-related hospital costs; instead, these costs increased, which has been attributed to efforts by hospitals to compensate for their financial losses. 285 Regarding mental health, the introduction of practice assistants or nurses in primary care from 2008 has yet to be fully implemented; people with mild mental health problems continue to be treated by specialists and are not quickly referred back to primary care. 286 Empirical evidence on the effectiveness of new professionals such as skin therapists in reducing referrals to dermatologists is currently lacking,287 although the introduction of optometrists or ophthalmic technicians has been linked to a reduction in referrals to eye specialists. 288

Search strategy

1. primary health care/

2. family practice/

3. physicians, family/

4. primary care.ti,ab.

5. primary health care.ti,ab.

6. general practice$.ti,ab.

7. family practice$.ti,ab.

8. general practitioner$.ti,ab.

9. (gp or gps).ti,ab.

10. (family adj (physician$ or doctor$)).ti,ab.

11. primary care practitioner$.ti,ab.

12. or/1-11

13. gatekeeping/

14. gpwsi$.ti,ab.

15. ((gp$ or practitioner$) adj2 special adj2 interest$).ti,ab.

16. ((outreach or specialist$ or satellite) adj clinic$).ti,ab.

17. (liaison adj3 (service$ or provid$ or provision or organis$ or organiz$ or deliver$ or attachment$)).ti,ab.

18. gatekeep$.ti,ab.

19. shared care$.ti,ab.

20. (integrated adj2 care).ti,ab.

21. (discharge adj (guideline$ or procedure$ or arrangement$ or routine$)).ti,ab.

22. ((primary or gp or gps) adj3 secondary care).ti,ab.

23. ((practice$ or practitioner$) adj3 incentive$).ti,ab.

24. (‘model of care’ or ‘models of care’).ti,ab.

25. “referral management”.mp.

26. exp Health Care Rationing/ or “health care ration*”.mp.

27. or/13-26

28. 12 and 27

29. limit 28 to (english language and yr=“Feb 2005 - current”)

30. outpatients/

31. ambulatory care facilities/

32. pain clinics/

33. surgicenters/

34. exp ambulatory care/

35. outpatient clinics, hospital/

36. ambulatory surgical procedures/

37. surgical procedures, minor/

38. outpatient$.ti,ab.

39. day surgery.ti,ab.

40. day case surgery.ti,ab.

41. day care surgery.ti,ab.

42. exp Telemedicine/ or telemonitor*.mp.

43. exp Remote Consultation/ or “remote consult*”.mp.

44. or/30-43

45. ‘referral.mp. and consultation’/

46. referral$.ti,ab.

47. waiting lists/

48. (waiting adj (time or times or list or lists)).ti,ab.

49. appointment$.ti,ab.

50. patient admission/

51. exp “Appointments and Schedules”/

52. admission$.ti,ab.

53. or/45-52

54. 12 and 44 and 53

55. limit 54 to (english language and yr=“Feb 2005 - current”)

56. “decision making”.mp. or exp Decision Making/

57. “decision support”.mp.

58. exp Choice Behavior/ or choice.mp.

59. “choose and book”.mp.

60. “choose & book”.mp.

61. exp “Appointments and Schedules”/ or appointment*.mp.

62. exp Decision Support Techniques/

63. exp Patient Preference/

64. exp Patient Participation/

65. “health litera$”.mp.

66. exp Health Literacy/

67. exp Patient Education as Topic/

68. exp Consumer Health Information/

69. “shared decision making”.mp.

70. or/56-69

71. 12 and 53 and 70

72. limit 71 to (english language and yr=“Feb 2005 - current”)

73. 29 or 55 or 72

Search strategy

1. primary health care/

2. family practice/

3. physicians, family/

4. primary care.ti,ab.

5. primary health care.ti,ab.

6. general practice$.ti,ab.

7. family practice$.ti,ab.

8. general practitioner$.ti,ab.

9. (gp or gps).ti,ab.

10. (family adj (physician$ or doctor$)).ti,ab.

11. primary care practitioner$.ti,ab.

12. or/1-11

13. gatekeeping/

14. gpwsi$.ti,ab.

15. ((gp$ or practitioner$) adj2 special adj2 interest$).ti,ab.

16. ((outreach or specialist$ or satellite) adj clinic$).ti,ab.

17. (liaison adj3 (service$ or provid$ or provision or organis$ or organiz$ or deliver$ or attachment$)).ti,ab.

18. gatekeep$.ti,ab.

19. shared care$.ti,ab.

20. (integrated adj2 care).ti,ab.

21. (discharge adj (guideline$ or procedure$ or arrangement$ or routine$)).ti,ab.

22. ((primary or gp or gps) adj3 secondary care).ti,ab.

23. ((practice$ or practitioner$) adj3 incentive$).ti,ab.

24. (‘model of care’ or ‘models of care’).ti,ab.

25. “referral management”.mp.

26. exp Health Care Rationing/ or “health care ration*”.mp.

27. or/13-26

28. 12 and 27

29. limit 28 to (english language and yr=“Feb 2005 - current”)

30. outpatients/

31. ambulatory care facilities/

32. pain clinics/

33. surgicenters/

34. exp ambulatory care/

35. outpatient clinics, hospital/

36. ambulatory surgical procedures/

37. surgical procedures, minor/

38. outpatient$.ti,ab.

39. day surgery.ti,ab.

40. day case surgery.ti,ab.

41. day care surgery.ti,ab.

42. exp Telemedicine/ or telemonitor*.mp.

43. exp Remote Consultation/ or “remote consult*”.mp.

44. or/30-43

45. ‘referral.mp. and consultation’/

46. referral$.ti,ab.

47. waiting lists/

48. (waiting adj (time or times or list or lists)).ti,ab.

49. appointment$.ti,ab.

50. patient admission/

51. exp “Appointments and Schedules”/

52. admission$.ti,ab.

53. or/45-52

54. 12 and 44 and 53

55. limit 54 to (english language and yr=“Feb 2005 - current”)

56. “decision making”.mp. or exp Decision Making/

57. “decision support”.mp.

58. exp Choice Behavior/ or choice.mp.

59. “choose and book”.mp.

60. “choose & book”.mp.

61. exp “Appointments and Schedules”/ or appointment*.mp.

62. exp Decision Support Techniques/

63. exp Patient Preference/

64. exp Patient Participation/

65. “health litera$”.mp.

66. exp Health Literacy/

67. exp Patient Education as Topic/

68. exp Consumer Health Information/

69. “shared decision making”.mp.

70. or/56-69

71. 12 and 53 and 70

72. limit 71 to (english language and yr=“Feb 2005 - current”)

73. 29 or 55 or 72

Search strategy

1. primary health care/

2. primary health care.ti,ab.

3. primary health care.ti,ab.

4. general practice$.ti,ab.

5. family practice$.ti,ab.

6. general practitioner$.ti,ab.

7. (gp or gps).ti,ab.

8. (family adj (physician$ or doctor$)).ti,ab.

9. primary care practitioner$.ti,ab.

10. or/1-9

11. gatekeeping/

12. gatekeep$.ti,ab

13. gpwsi$.ti,ab.

14. ((gp$ or practitioner$) adj2 special adj2 interest$).ti,ab.

15. ((outreach or specialist$ or satellite) adj clinic$).ti,ab.

16. (liaison adj3 (service$ or provid$ or provision or organis$ or organiz$ or deliver$ or attachment$)).ti,ab.

17. shared care$.ti,ab.

18. (integrated adj2 care).ti,ab.

19. (discharge adj (guideline$ or procedure$ or arrangement$ or routine$)).ti,ab.

20. ((primary or gp or gps) adj3 secondary care).ti,ab.

21. ((practice$ or practitioner$) adj3 incentive$).ti,ab.

22. (‘model of care’ or ‘models of care’).ti,ab.

23. “referral management”.mp.

24. exp Health Care Rationing/ or “health care ration*”.mp.

25. or/11-24

26. 10 and 25

27. outpatients/

28. ambulatory care facilities/

29. pain clinics/

30. surgicenters/

31. exp ambulatory care/

32. outpatient clinics/

33. ambulatory surgical procedures/

34. outpatient$.ti,ab.

35. day surgery.ti,ab.

36. day case surgery.ti,ab.

37. day care surgery.ti,ab.

38. exp Telemedicine/ or telemonitor*.mp.

39. exp Remote Consultation/ or “remote consult*”.mp.

40. or/27-39

41. ‘referral.mp. and consultation’/

42. referral$.ti,ab.

43. waiting lists/

44. (waiting adj (time or times or list or lists)).ti,ab.

45. appointment$.ti,ab.

46. appointment*.mp.

47. patient admission/

48. admission$.ti,ab.

49. or/41-48

50. 10 and 40 and 49

51. “decision support”.mp.

52. “choose and book”.mp.

53. “choose & book”.mp.

54. exp Patient Participation/

55. “health litera$”.mp.

56. exp Health Literacy/

57. exp Consumer Health Information/

58. “shared decision making”.mp.

59. choice.mp.

60. or/51-60

61. 10 and 49 and 61

62. 26 or 50 or 62

63. limit 63 to yr=“Feb 2005 –Current”

Search strategy