|Home | About | Journals | Submit | Contact Us | Français|
Objective To review trials of nurse led interventions for hypertension in primary care to clarify the evidence base, establish whether nurse prescribing is an important intervention, and identify areas requiring further study.
Design Systematic review and meta-analysis.
Data sources Ovid Medline, Cochrane Central Register of Controlled Trials, British Nursing Index, Cinahl, Embase, Database of Abstracts of Reviews of Effects, and the NHS Economic Evaluation Database.
Study selection Randomised controlled trials of nursing interventions for hypertension compared with usual care in adults.
Data extraction Systolic and diastolic blood pressure, percentages reaching target blood pressure, and percentages taking antihypertensive drugs. Intervention effects were calculated as relative risks or weighted mean differences, as appropriate, and sensitivity analysis by study quality was undertaken.
Data synthesis Compared with usual care, interventions that included a stepped treatment algorithm showed greater reductions in systolic blood pressure (weighted mean difference −8.2 mm Hg, 95% confidence interval −11.5 to −4.9), nurse prescribing showed greater reductions in blood pressure (systolic −8.9 mm Hg, −12.5 to −5.3 and diastolic −4.0 mm Hg, −5.3 to −2.7), telephone monitoring showed higher achievement of blood pressure targets (relative risk 1.24, 95% confidence interval 1.08 to 1.43), and community monitoring showed greater reductions in blood pressure (weighted mean difference, systolic −4.8 mm Hg, 95% confidence interval −7.0 to −2.7 and diastolic −3.5 mm Hg, −4.5 to −2.5).
Conclusions Nurse led interventions for hypertension require an algorithm to structure care. Evidence was found of improved outcomes with nurse prescribers from non-UK healthcare settings. Good quality evidence from UK primary health care is insufficient to support widespread employment of nurses in the management of hypertension within such healthcare systems.
Essential hypertension is a major cause of cardiovascular morbidity.1 In 2003 the prevalence of hypertension in England was 32% in men and 30% in women.2 Since the prevalence of hypertension increases with age it is a growing public health problem in the Western world faced with ageing populations.3 The lowering of raised blood pressure in drug trials has been associated with a reduction in stroke of 35-40%, heart attack of 20-25%, and heart failure of over 50%.4 To achieve these benefits, aggressive and organised treatment to attain blood pressure targets is required, yet often contacts with health professionals do not trigger changes in antihypertensive therapy5; a phenomenon termed “clinical inertia.”6
Most patients require a combination of antihypertensive drugs to reach target blood pressure. Guidelines advocate logical drug combinations,7 and in England the National Institute for Health and Clinical Excellence has published a treatment algorithm for clinicians to follow.8 Hypertension is a condition almost entirely managed in primary care, and in the United Kingdom is an important component of the Quality and Outcomes Framework, which rewards practices for achievement of blood pressure standards set by the National Institute for Health and Clinical Excellence.9 Achievement between practices, however, varies considerably10 and knowledge of guidelines among general practitioners does not necessarily translate into their implementation.11
Doubt persists about how best to organise effective care and interventions to control hypertension by the primary care team. In 2005 a Cochrane review classified 56 trials of interventions into six categories: self monitoring, education of patients, education of health professionals, care led by health professionals (nurses or pharmacists), appointment reminder systems, and organisational interventions. The review concluded that an organised system of regular review allied to vigorous antihypertensive drug therapy significantly reduced blood pressure and that a stepped care approach for those with blood pressure above target was needed.12 Nurse or pharmacist led care was suggested to be a promising way forward but required further evaluation. Another review found that appropriately trained nurses can produce high quality care and good health outcomes for patients, equivalent to that achieved by doctors, with higher levels of patient satisfaction.13 Nurse led care is attractive as it has been associated with stricter adherence to protocols, improved prescribing in concordance with guidelines, more regular follow-up, and potentially lower healthcare costs. Without associated changes in models of prescribing, however, there seems to be little effect on blood pressure level.14 At present the usual model of care is shared between general practitioners and practice nurses, with general practitioners prescribing. Our local survey of Devon and Somerset found that of 79 responding practices (n=182; response rate 43%) 53 were using this model, with only four using nurse led care, including nurse prescribing (unpublished observation). In the light of these uncertainties over models of care and whether blood pressure reduction with nurse led care can be achieved, we explored further the trial evidence for efficacy of nurse led interventions through a systematic review. To elucidate whether nurse prescribing is an important component of this complex intervention and to identify areas in need of further study, we reviewed the international evidence base for such an intervention and its applicability to primary care in the United Kingdom.
We searched the published literature for randomised controlled trials that included an intervention delivered by nurses, nurse prescribers, or nurse practitioners designed to improve blood pressure, compared with usual care. The population of interest was adults aged 18 or over with newly diagnosed or established hypertension above the study target, irrespective of whether or not they were taking antihypertensive drugs. Primary outcome measures were systolic and diastolic blood pressure at the end of the study, changes in systolic and diastolic blood pressure compared with baseline, percentage of patients reaching target blood pressure, and percentage taking antihypertensive drugs. The secondary outcome was cost or cost effectiveness of interventions.
We searched Ovid Medline, the Cochrane Central Register of Controlled Trials, British Nursing Index, Cinahl, Embase, Database of Abstracts of Reviews of Effects, and the NHS Economic Evaluation Database. Using a strategy modified from the previous review of 2005 we searched for randomised controlled trials in original English language and published between January 2003 and November 2009.12 We identified older citations from this review, hence the choice of cut-off date for the search (see web extra). We also corresponded with authors to identify missed citations.
Two authors (CEC, LFPS) independently selected potentially relevant studies by screening retrieved citations and abstracts. Trials assessed as definite or uncertain for inclusion were retrieved as full papers. Differences were resolved by discussion; arbitration from a third author (JLC) was planned but not required. Two authors (CEC, LFPS) independently extracted details of the studies and data using a standardised electronic form, with differences resolved by discussion. Risk of bias in the generation of the randomisation sequence, allocation concealment, and blinding (participants, carers, assessors) was assessed as adequate, uncertain, or inadequate using Cochrane criteria.15 One author (LFPS) checked the reference lists of all included studies for further potentially relevant citations, and two authors (CEC, LFPS) reviewed this list and agreed on further potentially relevant papers to retrieve in full. Searches were undertaken in June 2009 and repeated in November 2009 before final writing up.
Data were pooled and analysed using RevMan v5.0.16 We carried out separate analyses for each intervention and outcome measure compared with usual care. Intervention effects were calculated as relative risks with 95% confidence intervals for dichotomous data. For continuous data we used a conservative random effects meta-analysis model to calculate mean differences and weighted mean differences with 95% confidence intervals. When a study included more than one intervention group with a single comparator arm, we included both intervention groups and split the number of patients in the common comparator arm across the separate intervention arms.15 Where required we calculated standard deviations from standard errors or confidence intervals presented within papers. Heterogeneity was quantified using the I2 statistic and the χ2 test of heterogeneity. Using sensitivity analysis we explored heterogeneity by excluding single outlying results or restricting analysis to studies of good quality. We reported pooled data only when heterogeneity was not significant (P>0.05). Two authors (CEC, RST) reviewed the data from cluster randomised controlled trials and either extracted the data as presented when the authors were deemed to have taken account of cluster effects or first adjusted using a design factor,15 with intraclass correlation coefficients for systolic and diastolic blood pressure derived from cluster studies in primary care.17
Searches identified 1465 potential citations. A further 66 potential studies were identified from citations in retrieved papers. After initial screening of the titles and abstracts 71 full studies were assessed for possible inclusion in the review and 33 met the inclusion criteria (fig 11).
Table 11 summarises the characteristics of the included studies. Seven cluster randomised controlled trials were randomised at practice18 19 20 21 22 23 or family level.24 Five described adjustment for clustering effects but two did not seem to have done so, therefore these were adjusted for cluster size.23 24 One study used a two level nested design of interventions at provider and patient level; combined patient level outcomes were extracted where possible, or as separate intervention and control groups for each provider intervention.25 Four studies had three arms. Three compared telephone monitoring and face to face nurse monitoring with usual care26 27 28 and outcomes were extracted as separate groups; one compared nurse and general practitioner interventions with usual care and only the nurse and control outcomes were extracted.21 The remaining randomised controlled trials were two armed studies randomised at individual patient level.
Interventions were categorised as nurse support delivered by either telephone (seven studies),25 26 27 28 29 30 31 community monitoring (defined as home or other non-healthcare setting; eight studies),24 26 32 33 34 35 36 37 or nurse led clinics. These were held in either primary care (13 studies)20 21 22 23 27 28 35 38 39 40 41 42 43 or secondary care (six studies).44 45 46 47 48 49 One study used alternate sessions with nurses at home and in general practice.50 Fourteen studies included a stepped treatment algorithm18 19 21 22 23 24 30 31 35 37 38 40 47 48 and nine included nurse prescribing in their protocol.24 30 31 35 37 40 44 47 48
Although most of the studies recruited participants with hypertension, 11 also recruited participants with diabetes,18 19 22 23 31 36 37 44 46 47 48 five with coronary heart disease,20 21 33 39 50 and one the siblings of patients with coronary heart disease.24 Most studies recruited predominantly white participants. Four studied hypertension care provided to African Americans,24 26 29 40 three to Chinese,33 34 46 two to South Asians,19 23 one to American Indians,37 and two to mixed non-white populations.44 45 Thirty eight studies were excluded after review of the full paper (fig 1).
Overall, study quality was moderate; random sequence generation was adequate in 70% (23/33) of studies, allocation concealment in 58% (19/33), and blinding of data collection in 43% (14/33); one study was described as an open (unblinded) randomised controlled trial.41 Thirteen studies were assessed as adequate in two of the three domains and adequate or unclear for the third.20 22 25 26 29 30 32 33 34 40 42 46 48 These studies were defined as of “good quality” and were used for sensitivity analysis by study quality. Only three of these reported UK trials; one of patients with ischaemic heart disease and hypertension20 and two of people with diabetes and hypertension.22 48 The method of blood pressure measurement was not described in 12 studies,19 20 21 22 23 24 33 39 42 43 46 47 10 used automated monitors,18 26 27 28 29 30 36 37 44 48 and seven referred to authoritative guidelines for measurement.25 32 34 41 44 45 50
Pooling of data across different types of interventions was limited by noticeable statistical heterogeneity between studies, which was not explained by restriction to good quality studies. Consequently the results are presented as subgroup analyses by type of intervention (table 22).). (See web extra for forest plots for all comparisons; summary statistics were omitted if significant heterogeneity was present; see table 2). One study did not report any estimates of variance and did not contribute data to the meta-analyses.42
Fourteen studies included a stepped treatment algorithm in their intervention18 19 21 22 23 24 30 31 35 37 38 40 47 48 and for nine it was the main focus of the intervention.18 19 21 22 35 37 38 47 48 Two studies of good quality30 40 showed greater magnitudes of reductions in blood pressure with the use of an algorithm compared with usual care: weighted mean difference, systolic −9.7 mm Hg (95% confidence interval −14.0. to −5.4 mm Hg) and diastolic −4.3 mm Hg (−7.4 to −1.2 mm Hg). Pooling of all four studies also showed a greater magnitude of reduction in systolic blood pressure (−8.2 mm Hg, −11.5 to −4.9; fig 22)23 30 37 40 with the use of an algorithm compared with usual care.
Pooling of three good quality studies22 40 48 showed no significant difference in achievement of study blood pressure targets in favour of an intervention including an algorithm (relative risk 1.09, 95% confidence interval 0.93 to 1.27). Although a total of 10 studies reported this outcome,18 19 22 31 35 38 40 42 47 48 statistical and clinical heterogeneity between them was significant.
Nine studies included nurse prescribing in their protocol; three in secondary care settings,44 47 48 three using community interventions,24 35 37 two using telephone monitoring,30 31 and one based in primary care.40
Two good quality studies30 40 showed greater magnitudes of blood pressure reductions for nurse prescribing than for usual care: weighted mean difference, systolic −9.7 mm Hg (95% confidence interval −14.0 to −5.4) and diastolic −4.3 mm Hg (−7.4 to −1.2). Pooling of all studies showed similar reductions: systolic −8.9 mm Hg (−12.5 to −5.3) and diastolic −4.0 mm Hg (−5.3 to −2.7; fig 33).
Only one good quality study reported absolute blood pressure as an outcome, but pooling of four studies showed a significantly lower absolute outcome systolic blood pressure in favour of nurse prescribing: weighted mean difference −7.2 mm Hg (95% confidence interval −10.9 to −3.5).30 31 37 40
Two good quality studies showed no difference in achievement of study blood pressure target (relative risk 1.20, 95% confidence interval 0.96 to 1.50).40 48 Significant statistical and clinical heterogeneity precluded further pooled analysis.
Seven studies included telephone monitoring of blood pressure by nurses.25 26 27 28 29 30 31 Meta-analysis of four groups from three good quality studies showed no significant difference in outcome systolic blood pressure (weighted mean difference −2.9 mm Hg, 95% confidence interval −7.5 to 1.6).25 26 29 Pooling of all studies gave a similar result (−3.5 mm Hg, −7.4 to 0.4; fig 44),), and pooling of three studies also showed no difference for outcome diastolic blood pressure (−1.1 mm Hg, −5.8 to 3.6).26 29 31
Pooled data from three studies25 27 31 (one of good quality25) showed a higher achievement of study blood pressure targets with telephone monitoring than with usual care (relative risk 1.24, 95% confidence interval 1.08 to 1.43).
Eight studies involved nurse interventions delivered outside of healthcare settings. Locations included the home,32 33 37 50 community centres,24 26 or a choice of both.34 One study was set in the workplace35 and one in a pharmacy.36 Pooled data from four good quality studies26 32 33 34 showed a lower outcome systolic blood pressure in favour of monitoring in the community (weighted mean difference −3.4 mm Hg, 95% confidence interval −6.1 to −0.7; fig 55)) and two good quality studies showed greater magnitudes of blood pressure reduction with community monitoring than with usual care: systolic −4.7 mm Hg (−8.3 to −1.2) and diastolic −3.1 mm Hg (−4.8 to −1.3).32 34 Pooling of data from all four studies also showed a greater magnitude of reductions in favour of the intervention: systolic −4.8 mm Hg (−7.0 to −2.7)32 34 36 37 and diastolic −3.5 mm Hg (−4.5 to −2.5).32 34 35 37
Four studies,32 35 36 50 including one of good quality,32 reported significantly better achievement of blood pressure targets in favour of the intervention, but significant heterogeneity precluded pooled analysis.
Fourteen studies were of nurse led clinics in primary care20 21 22 23 27 28 35 38 39 40 41 42 43 50 and six in secondary care settings.44 45 46 47 48 49 For primary care studies, two of good quality showed no difference in diastolic blood pressure (−2.9 mm Hg, −6.9 to 1.1).20 40 Pooling of all studies showed a greater magnitude of reduction in blood pressure for nurse led clinics compared with usual care (systolic −3.5 mm Hg, −5.9 to −1.1 and diastolic −1.9 mm Hg, −3.4 to −0.5; fig 66),23 27 28 40 41 and two good quality studies showed no difference in achievement of blood pressure targets with nurse led clinics (relative risk 1.14, 95% confidence interval 0.83 to 1.57).22 40
For secondary care clinics, only two were of good quality and did not report comparable outcomes.46 48 For all studies, pooling of data from three studies showed no difference in outcome diastolic blood pressure (weighted mean difference −1.4 mm Hg, −3.6 to 0.86)44 46 49 and no greater achievement of study blood pressure targets (relative risk 1.47, 95% confidence interval 0.79 to 2.74)44 47 48 in nurse led clinics compared with usual care.
Significantly lower systolic blood pressure was achieved for any nurse led intervention for four groups from three good quality studies recruiting African American participants (weighted mean difference −7.8 mm Hg, 95% confidence interval −14.6 to −0.9) 24 29 40 but neither systolic nor diastolic blood pressure was lower on pooling of three good quality studies of Chinese participants (systolic −2.6 mm Hg, −7.5 to 2.3 and diastolic −0.5 mm Hg, −2.3 to 1.3; fig 77).33 34 46 Pooling of two studies, neither of good quality, showed no significant increase in the use of antihypertensive drugs in South Asian participants (relative risk 1.22, 95% confidence interval 0.90 to 1.65),19 23 but pooling of four studies across different ethnic groupings did show a small increase in favour of any nurse led intervention compared with usual care (1.22, 1.02 to 1.47).19 23 24 44
Only four studies presented any data. From the United Kingdom one study reported a cost per patient of £434 (€525, $632) over two years to provide additional nurse clinics and support from specialist nurses, representing £28933 per quality adjusted life year gained19 and another study found that primary care costs were £9.50 per patient compared with £5.08 for usual care.43 In the United States a study reported a 50% higher total cost of staff at $134.68 (£92.65, €111.90) per patient treated in a nurse led clinic compared with $93.70 for usual care,47 but a Mexican study reported $4 (£2.75, €3.32) per patient or $1 per 1 mm Hg reduction of systolic blood pressure.32
In comparison with usual patterns of care, nurse led interventions that included a stepped treatment algorithm showed significantly greater reductions of systolic and diastolic blood pressure, but this was not associated with higher achievement of blood pressure targets. Studies incorporating nurse led prescribing also showed bigger reductions of systolic and diastolic blood pressure. Telephone monitoring was associated with higher achievement of study targets for blood pressure. Community monitoring showed lower outcome systolic blood pressure, greater reductions in systolic and diastolic blood pressure, and, although pooling of data was not possible, greater achievement of study blood pressure targets. Nurse led clinics in primary care achieved greater reductions in systolic and diastolic blood pressure compared with usual care. No clear beneficial effects on our primary outcomes were observed from secondary care clinics.
Pooled interventions showed significantly lower systolic blood pressure in African American participants with nurse led interventions than with usual care, but little difference for other ethnic minority groups.
Since blood pressure was reported variously as final blood pressure or change from baseline for systolic or diastolic readings, less pooling of results was possible than may have been anticipated.
Thirteen of the 33 included randomised controlled trials met our quality criteria. Only three of these were from the United Kingdom20 22 48 and none investigated an unselected primary care hypertensive population. Therefore the evidence base for nurse led care of hypertension in the United Kingdom relies on generalisation of findings from other, principally American, healthcare systems. In total, 12 trials were identified from the United Kingdom, of which six studied blood pressure control in people with diabetes18 19 22 23 44 48, four in patients with ischaemic heart disease,20 21 39 50 and two in people with uncontrolled hypertension.38 43
We restricted our search to articles in English, which may have excluded some potential international data; however, we consider it unlikely that significant evidence applicable to UK health care would have only been published in another language.
The usual reason for judging a trial’s quality as inadequate was weakness of blinding. As it was not possible for the participants to always be blinded to whether they were seeing a doctor, nurse, or other health professional, this limitation must be accepted for any face to face intervention. We aimed to assess blinding of the researchers collecting outcome data to the intervention; these were often the same nurses who delivered the intervention and therefore were open to bias. This lack of formal blinding in trials is recognised as a methodological challenge51 but need not be seen as a limitation because implementation of these findings would also necessarily be unblinded, so a pragmatic approach to studying these interventions can be relevant.52 Future trials will, however, need careful design to minimise bias.
One third of studies gave no description of the method used to measure blood pressure and only seven referred to published guidelines on blood pressure measurement, therefore the reliability of reported outcome measures cannot be judged easily.
Although interventions such as use of algorithms and nurse prescribing were associated with meaningful blood pressure reductions there was not a concomitant rise in achievement of target blood pressure. Although apparently inconsistent this could be a sample size effect, with some studies underpowered to show differences in dichotomous outcomes. It may also be explained by the noticeable variation in individual blood pressure targets in the studies, which were sometimes composite or multiple.18 19 35 44 Therefore reporting of absolute blood pressure reductions may be the more robust outcome measure for comparison in future reviews.
Many studies combined the use of a treatment algorithm with the nurse intervention; therefore the results contributed to both analyses. It was not possible within this review to separate out thoroughly the components of each intervention that were or were not effective.
For most studies the duration of follow-up was relatively short; only five followed participants for more than 12 months.19 21 25 40 41 Therefore it is not possible to extrapolate the findings as sustained benefits of the interventions.
We present evidence of benefit in some studies of ethnic minority groups because hypertension is recognised to carry higher levels of morbidity and mortality in some such populations.8 These findings, however, pool different types of intervention so cannot identify specific nurse led interventions of benefit in these groups. Furthermore, the “usual care” arm of some studies, predominantly from America,24 26 29 40 represented minimal care; therefore the benefits shown may be larger than could be expected if introduced to more inclusive healthcare systems, such as are found in the United Kingdom.
We included cost and cost effectiveness as a secondary outcome measure. It is, however, possible that other papers discussing this outcome (that is, non-randomised controlled trials) were not retrieved by our search strategy. Therefore a more thorough primary review of cost data may be needed.
The traditional view of the nurse’s role in hypertension care is to educate, advise, measure blood pressure,51 and enhance self management.53 Previous reviews have suggested that nurse led care may achieve better outcomes by increased adherence to protocols and guidelines, but we found insufficient evidence to confirm this.14 The most recent review12 identified an organised system of regular review and stepped care as essential components of successful interventions. This updated review supports this view, showing benefits in blood pressure reduction with the use of a treatment algorithm. No previous review has found sufficient evidence to support the assertion that nurse prescribing should be a key component of nurse interventions for hypertension14; however, this review has shown better blood pressure outcomes in favour of nurse prescribing based on studies in American healthcare systems.
Interventions varied greatly in intensity and presumably therefore in cost. Lack of information on cost effectiveness has been identified previously,54 and although this was only a secondary outcome measure for this review we noted that only four studies, including one of good quality,32 reported on costs.19 32 43 47 All four showed higher costs for the intervention, approaching 50% higher in two cases.43 47 Only one study seemed to be cost effective,32 but costs depend on the healthcare system within which the intervention is delivered, so we were unable to show any cost benefit that could be generalised across differing systems. Although nurses may save on salary costs, the evidence is conflicting, with potential savings being offset by an increased length of consultation.55 Evidence of cost benefit in acute self limiting conditions56 cannot be assumed to translate to the management of chronic disease, so future trials should incorporate a formal cost effectiveness analysis within their design.
Hypertension is identified with higher prevalence and morbidity levels in some ethnic minority groups such as African Americans and South Asians.57 Studies recruiting from these populations found significant reductions in blood pressures with any nurse led intervention. For studies from non-UK healthcare systems, “usual care” represented minimal or absent care.29 40 We therefore interpret this with caution.
The delivery of nurse led care in chronic conditions is a complex intervention. This review suggests that such care can improve on doctor led or usual care of hypertension. The key component of an intervention seems to be a structured treatment algorithm, and we have found evidence in favour of nurse prescribing. Although no clear benefits were seen for secondary care clinics improvements were found in both primary care and community based settings, suggesting that these findings can be applied to primary care clinics in the United Kingdom, or equivalent community settings in other healthcare systems. Although the absolute differences in blood pressure seem small—for example, a 4 mm Hg greater reduction in diastolic blood pressure with nurse prescribing than with usual care, a 2 mm Hg reduction in diastolic blood pressure is associated with a 15% reduction in risk of stroke or transient ischaemic attack in primary prevention.58 Similarly a 20-30% reduction in frequency of stroke, coronary heart disease, major cardiovascular events, and cardiovascular death is seen with a 3 mm Hg reduction in systolic blood pressure,59 and differences of this magnitude or greater are seen with nurse led clinics, nurse prescribing, and the use of an algorithm.
In this review we found international evidence of benefit from nurse led interventions but no evidence of good quality was derived from an unselected UK population with hypertension in primary care. Evidence from other healthcare systems cannot necessarily be generalised, therefore further studies relevant to the United Kingdom are needed. Such studies should ideally include a structured algorithm, examine the role of nurse led prescribing, and include a robust economic assessment. They should report absolute measures of blood pressure as this would best permit comparison with the existing literature and take care to minimise bias by blinding outcome assessors to the intervention.
Nurse led interventions for hypertension in primary care should include an algorithm to structure care and can deliver greater blood pressure reductions than usual care. There is some evidence of improved outcomes with nurse prescribers, but there is no evidence of good quality from United Kingdom studies of essential hypertension in primary care. Therefore, although this review has found evidence of benefit for nurse led interventions in the management of blood pressure, evidence is insufficient to support the widespread use of nurses in hypertension management within the UK healthcare systems.
We thank Kate Quinlan (East Somerset Research Consortium) for carrying out the searches and retrieving articles, Joy Choules (Primary Care Research Group) for helping retrieve articles, and Liam Glynn (Cochrane Hypertension Group) for sharing citation lists.
Contributors: CEC and LFP reviewed the literature search results, identified papers for retrieval, reviewed full papers for inclusion, and extracted data for meta-analysis. CEC and RST undertook the meta-analysis. JLC acted as study supervisor. All authors contributed to the interpretation of the findings and drafting of the manuscript. CEC is guarantor for the study.
Funding: This research was supported by the Scientific Foundation Board of the Royal College of General Practitioners and by the South West GP Trust.
Competing interests: All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any company for the submitted work; no financial relationships with any companies that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work.
Ethical approval: Not required.
Data sharing: No additional data available.
Cite this as: BMJ 2010;341:c3995