BACKGROUND: Chronically ill and ageing populations demand increasing human resources who can provide on-going and frequent follow-up care. We performed a systematic review to assess the effect of physician-nurse substitution on process care outcomes.

METHODS: We searched OVID Medline, Embase, CINAHL and The Cochrane Library for all available dates up to August 2012 and updated in February 2014. We selected and critically appraised published randomised controlled trials (RCT) and followed the PRISMA guidelines for the reporting of systematic reviews.

RESULTS: A total of 14 RCTs comprising 10,743 participants met the inclusion criteria. Studies were generally small and suffered from attrition of ≥20% and selection biases. There were 53 process measurements investigated in the 14 RCTs, many of which were unique to specific conditions. Accounts of nurses’ roles, responsibilities, tasks, qualifications and training content/components were not described in sufficient detail. Most study estimates showed no significant differences between nurse-led care and physician-led care while less than a half (~40%) favoured nurse-led care.

CONCLUSIONS: Despite the methodological limitations and the varying nurses’ roles and competencies across studies, specially trained nurses can provide care that is at least as equivalent to care provided by physicians for the management of chronic diseases, in terms of process of care. Future, larger studies with better quality methods are needed and should report and assess whether the differences in effects vary due to diversity in roles, qualifications, training competencies and characteristics of clinicians delivering substitution of care.

Key words: systematic review; physician-nurse substitution; primary healthcare; chronic disease; delivery of healthcare; process indicators

Introduction

The need for healthcare services is becoming more heterogeneous as the burden of chronic diseases and population ageing spreads rapidly. Chronically ill and ageing populations are expected to rise as the main users of healthcare services until at least mid-twenty-first century. Consequently, these populations are bound to have more extended care needs (including psychosocial and behavioural support) and an increased demand for human resources. A WHO report however, showed the global number of health care providers, namely physicians, nurses and midwives, remains lower than required per 1,000 population [1–3]. By 2011, 57 countries were still facing a critical shortage of these health workers including regions in Africa, Eastern Mediterranean, South-East Asia, Western Pacific, Americas and Europe [4]. Factors greatly contributing to this shortage include the low flow of primary care physicians (less graduates selecting and remaining in the primary care sector), changes in the working culture (e.g. newer generations of physicians working fewer hours than prior generations) and trends in retirement [5]. These changes have provided plenty of opportunities to create innovative staff-mix models [6].

It has long been suggested that nurses should perform greater roles and be granted full practice [7]. Indeed, nurses already provide an increasingly important contribution in primary care and although decreasing in number, they are still one of the largest groups of qualified healthcare professionals and they are also less expensive than physicians. Therefore their role for substituting physicians has gained increasing interest from policymakers hoping to address workforce shortages and maldistribution of workload, while reducing costs, especially in the care for the chronically ill [2, 8]. However at this time, it has been difficult to demonstrate how best to integrate nurses in a substitution model of care. Especially with the variability in definitions of nurses' roles and the diversity of competencies among roles with the same name or differing roles and skills among healthcare systems, this integration may seem doubtful. Nevertheless, nurses’ education keeps evolving in order to adjust to new demands in healthcare and nurses continue to support physicians in many areas in many countries. Substitution may take place in a wide range of care settings and/or clinical areas. It involves the transfer of tasks which are traditionally from the domain of physicians, to nurses who then take autonomous or delegated responsibilities to deliver care.

In 2002 and 2005, two systematic reviews explored the substitution of physicians by nurses in primary care and concluded with no appreciable differences between nurse-led care and physician-led care in terms of health outcomes (patient satisfaction, quality of life, other morbidity), use of resources and healthcare costs [9, 10]. However, authors found insufficient data on process outcomes other than length of consultations, amount of prescriptions and investigations, consultations and referrals. The identified evidence also had methodological limitations and lacked clear reporting of nurses’ roles (qualifications and training) and on the use of guidelines. We performed a timely update of previous reviews [9, 10] with a focus on the process of care outcomes, other than those examined in the mentioned reviews, and on the type and degree of nurses’ competencies.

Methods

This study was part of a large systematic review and meta-analysis project designed to assess the evidence of physician-nurse substitution in primary care. The methodological procedures of this systematic review are similar to those employed in the reviews of this project reported elsewhere [11]. We developed a protocol prior to the commencement of the review and followed the PRISMA guidelines [12] for the reporting of systematic reviews and meta-analyses (additional file, table S1).

Study inclusion and exclusion criteria

We searched for peer reviewed randomised or quasi-randomised (e.g., controlled before-and-after studies, interrupted time series) controlled trials (RCTs) published in English. Studies were eligible if they fulfilled the following criteria: examined populations of all ages and all conditions including mental health and addiction restricted to primary care; assigned patients to nurse-led care (all nurse roles) or physician-led care (family physicians, paediatricians and geriatricians) based on a substitution model; and if care interventions had taken place in general practices, community or ambulatory care settings regardless of the recruitment sources. We further limited the inclusion of studies to the report of process of care outcomes, which meant measures that reflected whether clinicians performed certain steps in diagnostic and treatments that are considered to be state of the art based on guidelines or good clinical practice (e.g. specific examinations/diagnostic tests, compliance with monitoring schemes). Following a framework published by a Cochrane review [10], we excluded studies in which nurses supplemented (i.e. complemented or extended care) the work of physicians or in which the effect of nurse-led care could not be distinguished from collaborative teams. We excluded measures of quality of life, satisfaction, mortality, hospital admissions, progression of disease and other clinical parameters.

Study identification and search strategy

We comprehensively searched OVID Medline, Embase, CINAHL and The Cochrane Library which includes the Cochrane Effective Practice and Organization of Care Group (EPOC). The original search was supported by an expert librarian. All searches were first performed for all available dates until August 2012. The searches were not restricted by age, date or country and included terms for ‘primary care’, ‘skill mix’, ‘doctors’, ‘nurse’, ‘substitution’ (additional file, table S2). We identified additional studies by manual searching of the reference lists of included studies and relevant reviews. Both electronic and manual searches were updated in February 2014.

Assessment of study quality

In view of the continuing debate about scoring the quality of trials, discussed by Juni et al. (1999) [13], a composite score was not performed. We assessed the risk of bias of individual trials following established criteria [14, 15] and provide a description of the studies’ adequacy regarding each item and an overall judgment of the quality of evidence. We considered bias due to attrition of at least 20% to be of significant concern and adequate intention-to-treat (ITT) when authors analysed participants based on their original group allocation [16].

Selection and assessment of studies

Two authors independently screened titles and abstracts, assessed the full-text of eligible publications and the risk of bias of included studies. Differences were resolved in discussion or by consensus with another author.

Data extraction

Two authors independently conducted data abstraction using a-priori designed and standardised data collection forms. We extracted information on bibliographic details, settings and characteristics of populations, interventions (nurses’ training competency and role, type of care, whether nurses were granted full clinical autonomy and whether interventions were delivered following specific protocols or guidelines, length of follow-up) and outcomes. Using the description of interventions and qualifications reported by study authors we grouped nurses’ training and roles into: nurse practitioner (NP) or nurse practitioner (NP+) who took or already had -for the purposes of the study- higher degree courses or had a specialisation; registered nurse (RN) or licensed nurse (LN). We also extracted quantitative and semi-quantitative data in dichotomous and/or continuous format. Data from trials reported across more than one publication were extracted as one study. If trials reported more than one comparison group of interest (e.g. family physicians and paediatricians), data were combined and compared as one to nurse-led care. We did not contact study authors to obtain additional information or data. Differences were resolved through consensus.

Statistical analyses

There was mostly one study per outcome thus we did not perform meta-analyses nor did we pre-specify subgroup analyses by clinical or methodological (risk of bias) characteristics. Where data were sufficiently reported, for each study outcome we calculated the unadjusted relative risks (RR) or the standard mean differences (SMD) and 95% confidence intervals (CI) of the absolute endpoints, using Review Manager (Version 5.1) [17]. We considered p <0.05 as statisticaly significant. The calculated effect sizes were tabulated with information on nurses’ roles and studies were arranged, within outcome categories, in increasing length of follow-up. The results were synthesised qualitatively. When scales pointed in opposite directions, we subtracted the mean from the maximum possible value of the scale or multiplied the mean of a set of studies by ‒1. We followed reported techniques to estimate standard deviations (SD) when these were missing [15]: using the information from the reported statistical analyses (e.g. from median and interquartile ranges), and if SDs of the final measurements were unavailable, we carried forward the baseline SDs assuming the intervention did not alter the variability of the outcome. Medians were treated differently from means and are reported distinguishably.

Results

Study identification

A total of 4,589 original records were identified by the electronic and manual searches. Of these, 44 publications were relevant, but 24 were excluded for the reasons provided in table S3 (additional file). Finally, 14 RCTs reported in 20 publications, met the inclusion criteria and comprised a total of 10,743 randomised participants [18–37]. Figure 1 shows the process of study selection.

fullscreen

Study and population characteristics

Table 1 and table S4 (additional file) show the study and population characteristics of included studies. All RCTs individually assigned participants to intervention and control groups and were conducted in the UK (n = 6), the Netherlands (n = 5), the USA (n = 2), and Russia (n = 1). Median follow-up was 12 (range 0.5 to 48) months with at least 12 months in 7 trials and less than 12 months in the other 7. The number of participants ranged from 60 to 1,859 with less than 200 (range 60 to 175) in six trials and more than 200 (range 230 to 1,859) in the other eight. Age was reported in twelve trials. Mean age ranged from 11.2 (SD 2.9) to 69.5 (SD 10.6) years in ten trials and age ranged from 0 to 83 in other two. In 13 RCTs that reported on gender, 49% of the population were male (range: 27 to 64%).

Settings and interventions

Table 2 and table S4 (additional file) report the characteristics of settings and interventions. Nurses worked as physician substitutes in physicians’ practices, nurse clinics, hospital outpatient clinics, reference clinics, and medical health centres. A total of 10 trials reported the number of participating nurses and/or physicians. In ten, the median number of nurses was 7.5 (range: 2 to 28) while in seven, the median number of physicians was 14 (range: 5 to 50). The location of practices (rural or urban) and social settings were scarcely reported. Nurses delivered care for a wide range of possible (diverse/undifferentiated/minor acute/common) or specific conditions (e.g. hypertension, heart failure, diabetes, HIV, etc.). Care provided ranged from single contact care, on-going care, first contact only care, first contact and on-going care, and first contact for urgent care. Only four trials [25, 27, 30, 36] reported the nurses’ years of experience (range: 0.17 to 12) and only one [25] reported that physicians had 16 years of experience. Nine trials employed NPs (NPs or NP+), and the rest employed LNs and/or RNs and/or NPs. Nurses were either enrolled as staff, took courses or already had professional qualifications by the time of participating in the study. Unfortunately, the educational preparation of nurses was reported insufficiently. Only four studies specifically stated that nurses had obtained an academic degree either a Masters in Advance Nursing [25], a Diploma in General Practice and implied an NP degree [30], a special degree in patient education [18], or had done a degree level course [36]. The terminology used by study authors to refer to participating nurses in the studies did not provide a straightforward definition of the various nurses’ educational degrees from the countries, at least for the UK, Netherlands and Russia. For example, a NP in the US requires a graduate degree whereas the UK did not seem to have a minimum educational requirement until this was recommended in 2012 for advanced NPs [38]. A “hypertension nurse” may then refer to a NP or a practice nurse both of whom are specialists in hypertension but each of whom might hold different educational degrees, for example basic education at diploma or degree level plus/or a bachelor’s or a master’s degree. The content of training or experience was often not described in detail. The lack of this information impeded a detailed assessment of the level of education and competencies, and the identification of common components across trials. All studies however, seem to assume that nurses fulfilled the appropriate clinical competency to deliver the study interventions. Responsibilities and tasks also varied across trials and were often incomplete. Nurses had full clinical autonomy in only two trials: one in patients with undifferentiated conditions [26] and one in diabetic patients [28]. In the remaining trials, nurses independently performed several tasks, but they still needed minor support from physicians, for example to report findings, sign prescriptions, referrals and hospitalisations, or to discuss patients’ records. Ten trials reported nurses’ interventions followed specific guidelines or protocols.

Risk of bias in the methods of the included studies

The overall quality of studies varied substantially when assessed against current reporting standards [14] (table 3). Inclusion and exclusion criteria were reported in 71% of the trials and funding sources in 64% of the trials. The success of the intervention was measured by defining a primary outcome in 50%. Among all trials, random sequence generation was adequate in 57%, allocation concealment in 50% and both criteria were adequate in 43%. No trial blinded both patients and providers. Patients were blinded in one trial, and outcome assessors were blinded in 36%. Patient or clinician crossover between groups was reported in two trials. Sample size calculation (80 to 90% power) was performed in 79% of the trials but only five maintained the required sample to achieve power. At baseline, patient groups were comparable for all tested factors in 71%. Attrition rate was ≥20% in 43%. Missing data (range 5 to 42%) was dealt with intention to treat (ITT) techniques in 29%.

Effectiveness of interventions

We identified 53 measurements of process of care reported in the 14 RCTs. Of these, 34 were reported in ten trials in which nurses cared for patients with specific conditions. The other 19 process of care measures were reported in four trials in which nurses cared for patients with more general conditions. Table 4 and table 5 show the individual trial estimates calculated from reported data.

Adherence to practical guidelines

Adherence of clinicians to practical guidelines was non-significant between groups in one trial [25].

Blood pressure management according to guidelines

Blood pressure management according to guidelines showed significantly more patients in the nurse group, compared to physicians, met a composite target (SBP and DPB, mm Hg: RR 1.1, 95% CI 1.06 to 1.13, p <0.00001) [20] and systolic blood pressure target (SBP, mm Hg RR 3.14, 95% CI 1.38 to 7.19, p = 0.007) [24]. However, the effect did not sustain at 48 [20] or 14 months [28] and was non-significant for DPB only [24].

Lipids

Significantly more patients in nurse-led care, compared to physicians, had appropriate secondary prevention of heart disease through the adequate management of lipids at 12 months following specific guidelines in one trial (RR 1.91, 95% CI 1.59 to 2.29, p <0.00001) [20]. However, these differences were non-significant at 14 or 48 months.

HbA1c

One trial showed no significant differences between nurse-led care and physician-led care in the number of patients who met the target values of glycated haemoglobin (HbA1c) according to guidelines for the management of diabetes mellitus type 2 at 14 months (HbA1c <7.0 or HbA1c ≤8.5) [28].

BMI and waist circumference

In one trial, compared to the physicians group, there were significantly more patients in the nurse group who had a decrease or regression in body mass index (BMI) at 6 months (RR 1.51, 95% CI 1.05 to 2.17, p = 0.03) [18] but the mean differences between groups at 12 or 14 months were non-significant in other two trials [28, 37]. One of these also showed significantly more patients in the nurse group who had a decrease or regression in waist circumference at 6 months (RR 2.36, 95% CI 1.34 to 4.16, p = 0.003) [18].

Asthma

Individual point estimates of one trial showed, compared to physician-led care, significantly more patients with a correct inhalation technique at 12 months in the nurse-led care group (RR 1.33, 95% CI 1.01 to 1.74, p = 0.04) [27], but the effect was non-significant at 24 months. In another trial, the differences between groups in the number of patients with well-controlled asthma were non-significant at 12 or 24 months [31].

Feet at risk

Compared to patients in physician-led care, there were significantly less patients with feet-not-at-risk in the nurse-led care group (of the patients who underwent measures to prevent the development of diabetic foot symptoms) (RR 0.59, 95% CI 0.42 to 0.82, p = 0.002) [28].

Visuomotor coordination

There was a significant improvement in best hand score in patients with nurse-led care, compared to patients in the physicians group in patients with Parkinson’s disease in one trial (MD –4.31, 95% CI –4.52 to –4.11, p <0.00001) [29].

Appropriate secondary prevention

Individual point estimates from one trial [20] showed significantly more patients with nurse-led care, compared to physician-led care, had appropriate secondary prevention of heart disease through the adequate management of aspirin intake (taken or contraindicated) (RR 1.22, 95% CI 1.14 to 1.31, p <0.00001) or low-fat diet (RR 1.16, 95% CI 1.03 to 1.31, p = 0.02) or moderate physical activity (RR 1.35, 95% CI 1.16 to 1.58, p = 0.0001) at 12 months. However, the effect did not sustain at 48 months except for moderate physical activity (RR 1.23, 95% CI 1.02 to 1.49, p = 0.03). Conversely, there were no significant differences between groups in the number of patients who had appropriate secondary prevention of heart disease through non-smoking at 12 or 48 months [20] or smoking cessation for asthma/COPD at 24 months in one trial [27].

Adherence to treatment, correct medication and diet

Adherence to treatment (in patients under anti-inflammatory agents) was non-significant between groups neither at 12 nor at 24 months in one trial [27]. Another trial showed a marginal significance that favoured the nurse-led care group in the number of patients who adhered to milk as part of their nutritional diet (RR 0.62, 95% CI 0.38 to 1, p = 0.05) [26] but no significant differences between groups in the number of patients who adhered to a special diet, bread taking, or with correct medication, at 6–12 months.

Patient information and knowledge

Seven trials reported on various types of provision of information, advice from clinicians or patient knowledge. There were generally one or two trials per outcome type. Nurses provided significantly more information than physicians on the causes of health problems or illness [22, 30] in two of the three trials with these data. Nurses also gave significantly more advice about relief of symptoms, duration of illness, how to reduce recurrences and what to do if problems persisted in one [30] of two trials [25, 30]; as well as more advice on self-medication and self-management [36], special exercises [26], and provided leaflets about patients illness [22]. Patients’ knowledge about the complications of disease was non-significant between groups [26]. In another trial, although data was collected, results were not reported [32].

There was no evidence of worsening outcomes with nurse-led care compared to physician-led care.

Discussion

Substitution of physicians by nurses has increased the possibility of achieving the quality outcomes required to satisfy the demands of an aging population and the shortages of physicians in many countries. We found however that the number of studies in this area is only increasing slowly despite previous reports [9, 10]. The studies were generally small and none fulfilled all quality criteria. More than 40% of the studies suffered from selection (lack/unclear allocation concealment and random generation) and attrition (≥20%) biases, and very probable publication bias since our review was limited to the published literature. Only a few studies maintained the sample required to achieve power and the length of follow up varied widely. It may not be surprising that most of the evaluated trials were conducted in Europe, mainly the UK, and the Netherlands. In the USA and Canada medical care has evolved to a shared role since NPs and physicians’ assistants were introduced or re-invigorated, both in the mid-1960s [39]. However, the USA is still in great need of more nurses to level out the shortage of primary care physicians [40, 41]. In the UK on the other hand, the introduction of nurses in advanced roles did not happen until early 1970s and the role of NPs were not consolidated until 1990 [42].

The studies generally assessed a large variety of process of care outcomes, sometimes with many unique measurements per study but only 38.3% (13/34) of those taken in patients with specific conditions and 42.1% (8/19) related to general conditions, significantly favoured nurse-led care compared to physicians. The remaining studies showed non-significant differences.

The competencies to treat the elderly and the chronically ill may differ among different types of nurses or physicians. Research has reported a reduction in mortality, failure-to-rescue rates, ulcers and length of stay after increasing the proportion of nurses who possess a bachelor’s degree [43–47]. However, unless stated by study authors, the evaluated literature does not offer sufficiently reported details on nurses’ educational level. The literature reflects an overuse of terminology for nurses’ job titles. In addition, nurses’ qualifications did not seem consistent among these, although nurses had received training to deliver the interventions. Nor was it possible to make clear judgments about nurses’ educational level by using a country’s definition of the terminology that authors used to refer to nurses. This is not surprising. In many countries, until recently a person did not have to hold a Bachelor’s degree in Nursing to be a nurse and nowadays nurses might have a masters or PhD degree. A nurse practitioner (NP) programme in the US for example requires a graduate degree [48]. In the UK, nurses working in NP roles were registered nurses who had undertaken a specific course of study to at least first degree (honours) level in 2008. By 2012, any nurses educationally prepared at bachelor’s or master’s level against the Royal College of Nursing competencies were entitled to be referred to as Advanced NPs [38]. This title requires a master’s degree in Wales but not necessarily in the other three countries of the UK. Whether nurses’ educational preparation or job titles (e.g. NP) with different educational degrees make a difference in the observed effects evaluated in this review cannot be concluded from the studies in question and should be examined further.

Accounts of responsibilities and tasks also varied across trials and were not described in sufficient detail. Due to the insufficient description of training content, we could not identify a common component across studies. In addition, it was generally assumed that nurses had the required competence to substitute physicians. However, the level of substitution (clinical autonomy) differed among trials and nurses seemed dependent of doctors’ supervision in most studies. This may suggest the importance of collaborative models of care. Research has shown that team approaches in which nurses, physicians and other clinicians work collaboratively, might lead to better outcomes [49]. The implementation of inter-professional care management programmes should also be considered in future research.

The use of process guidelines or protocols to deliver the interventions was reported in nearly 75% of the studies which suggests that adherence to treatment, diet and provider guidelines can result in nurse care better or similar to that of physicians. In fact, the use of evidence-based clinical guidelines has shown effective improvement in the process and structure of care [50], and has been reported to aid the transfer of tasks between clinicians while maintaining and improving the quality of care [51]. The lack of reporting in adherence to guidelines or protocols in the other 25.6% of the studies may just be an indicator of the level of adherence and availability of guidelines in practice. These differences may result from the time and method of development, type of health problem, content of recommendations, and source of dissemination within others [52]. The finding that nurses are significantly more likely than physicians to provide life-style advice and information about various aspects of disease is consistent with previous reviews [9, 10], research reporting positive associations of nurse lifestyle interventions in the prevention of chronic disease [53] and reports showing patients’ appreciation to nurses’ involvement especially in education and counselling [54].

Despite all limitations, no study showed harm of nurse-led care interventions compared to physician-led care. Trained nurses seem to provide equal or better care compared to physicians for the management of chronic disease through process of care measures, within their scope of practice. We speculate that, regardless of the healthcare system in which nurses substitute physicians, and given the heterogeneity in patient populations, settings and interventions, the reasons for these effects may be that specific components of nurse training and competency are shared among studies (e.g. patient education). Another possibility is that nurses may adhere to process care guidelines and protocols better than physicians. In addition, individual trial estimates suggest that the effects of interventions may only be significant at short term (≤12 months), for some conditions such as asthma and COPD. These factors may have a significant impact on the continuity and quality of patient care and should be investigated in future studies.

Strengths and limitations of this systematic review

To our knowledge, this is the first systematic review of physician-nurse substitution with a focus on process of care outcomes. Our review updates earlier systematic reviews [9, 10] and uses a comprehensive search of the literature and critical appraisal of RCTs which are at lower risk of bias than other study designs [14, 55]. A particular strength of our review is that we examined individual trial data in relation to nurses’ competencies and roles. It was however often difficult to understand in detail the role and responsibilities of nurses when substituting physicians. In many cases, nurses remain embedded in care teams that also involved physicians. It was also difficult to determine the nurses’ level of education and whether the training competencies were appropriate for the type of care delivered. In many cases the description of training content is insufficient and limited the identification of common program components across studies. We excluded studies in which nurse-led teams were compared to physician-led care in a primary care base because of the potential confounding in the type of care/tasks (e.g. specialised vs. routine) and the type of clinician delivering the interventions. A limitation of the literature is the small number of studies that met the inclusion criteria. Although we did not search for grey literature and included publications in English only, we used thorough electronic and hand searches including the screening of relevant reviews (some in foreign languages). We did not contact study authors to further obtain or clarify missing information.

Conclusion

Our systematic review suggests that, in terms of process of care outcomes, special trained nurses can provide care that is at least as equivalent to care provided by physicians for chronic diseases such as diabetes, cardiovascular disorders, asthma, COPD, and hypertension. One limiting factor is the small number of studies reporting many unique processes of care measures. It is unclear whether the observed effects are due to the diversity in nurses’ competencies, roles, and experiences. It is also unclear whether the components or contents of training competencies boost these effects. Future studies could benefit from the inclusion of larger samples, a more rigorous methodology, longer follow-up episodes, and mapping the wider range of nurse care from many countries. Consideration should be given to the role of multidisciplinary teams in which nurses embed their roles. Qualitative research could also add valuable information since it may allow the identification of factors that influence the performance and quality of care within the context of health care systems [56]. Future studies should especially provide precise accounts of the components of competencies or training programmes, and the qualifications, tasks and responsibilities of clinicians delivering substitution of care. In particular, the reporting of complex interventions according to recently proposed guidelines [57] may help establishing better reporting of substitution of care studies in the future. Consequently improving the interpretation of results and allowing the replication of interventions, so that future evaluations in decision-making can employ such evidence.

Annex: Supplementary data.

Table S1: PRISMA Checklist.

Table S2: Search strategy in Ovid Medline.

Similar search strategies were performed and run in EMBASE, The Cochrane Library of Systematic Reviews and CINAHL and include specific search filters for RCTs.

Table S3: Studies excluded with reasons for exclusion based on appraisal of full text articles.

Table S4: Characteristics of participants, interventions and outcomes in the studies included in review.

  1 World Health Organization. The world health report: working together for health. Geneva, Switzerland: World Health Organization, 2006.

  2 World Health Organization. Task shifting: rational redistribution of tasks among health workforce teams: global recommendations and guidelines. Geneva, Switzerland: World Health Organization, 2008.

  3 World Health Organization. World Health Statistics 2012: Indicator compendium. Geneva, Switzerland: World Health Organization, 2012.

  4 World Health Organization. Global Health Workforce Alliance: Reviewing Progress, Renewing Commitment. Progress Report on the Kampala Declaration and Agenda for Global Action. Geneva, Switzerland: World Health Organization, 2011.

  5 Salsberg E, Grover A. Physician workforce shortages: implications and issues for academic health centers and policymakers. Acad Med. 2006;81:782–7.

  6 Buchan J, O’May F. Determining skill mix in the health workforce: guidelines for managers and health professionals. Geneva, Switzerland: World Health Organization, 2000.

  7 De Geest S, Moons P, Callens B, Gut C, Lindpaintner L, Spirig R. Introducing Advanced Practice Nurses / Nurse Practitioners in health care systems: a framework for reflection and analysis. Swiss Med Wkly. 2008;138:621–8.

  8 Chopra M, Munro S, Lavis JN, Vist G, Bennett S. Effects of policy options for human resources for health: an analysis of systematic reviews. Lancet. 2008;371:668–74.

  9 Horrocks S, Anderson E, Salisbury C. Systematic review of whether nurse practitioners working in primary care can provide equivalent care to doctors. BMJ. 2002;324:819–23.

10 Laurant M, Reeves D, Hermens R, Braspenning J, Grol R, Sibbald B. Substitution of doctors by nurses in primary care. Cochrane Database Syst Rev 2005:CD001271

11 Martinez-Gonzalez NA, Tandjung R, Djalali S, Huber-Geismann F, Markun S, Rosemann T. Effects of physician-nurse substitution on clinical parameters: a systematic review and meta-analysis. PLoS One 2014;9:e89181.

12 Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gotzsche PC, Ioannidis JP, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLoS Med 2009;6:e1000100.

13 Juni P, Witschi A, Bloch R, Egger M. The hazards of scoring the quality of clinical trials for meta-analysis. JAMA. 1999;282(11):1054–60.

14 Higgins JP, Altman DG, Gotzsche PC, Juni P, Moher D, Oxman AD, et al. The Cochrane Collaboration’s tool for assessing risk of bias in randomized trials. BMJ. 2011;343:d5928.

15 Higgins JPT GS, (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0: The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org.

16 Fergusson D, Aaron SD, Guyatt G, Hebert P. Post-randomisation exclusions: the intention to treat principle and excluding patients from analysis. BMJ. 2002;325:652–4.

17 Review Manager (Revman) [Computer program]. Version 5.2.4. for Windows. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2012 [http://www.cochrane.org/resources/handbook/].

18 Andryukhin A, Frolova E, Vaes B, Degryse J. The impact of a nurse-led care programme on events and physical and psychosocial parameters in patients with heart failure with preserved ejection fraction: a randomized clinical trial in primary care in Russia. Eur J Gen Pract. 2010;16(4):205–14.

19 Campbell NC, Ritchie LD, Thain J, Deans HG, Rawles JM, Squair JL. Secondary prevention in coronary heart disease: a randomized trial of nurse led clinics in primary care. Heart. 1998;80(5):447–52.

20 Campbell NC, Thain J, Deans HG, Ritchie LD, Rawles JM. Secondary prevention in coronary heart disease: baseline survey of provision in general practice. BMJ. 1998;316(7142):1430–4.

21 Campbell NC, Thain J, Deans HG, Ritchie LD, Rawles JM, Squair JL. Secondary prevention clinics for coronary heart disease: randomized trial of effect on health. BMJ. 1998;316(7142):1434–7.

22 Chan D, Harris S, Roderick P, Brown D, Patel P. A randomized controlled trial of structured nurse-led outpatient clinic follow-up for dyspeptic patients after direct access gastroscopy. BMC Gastroenterol. 2009;9:12.

23 Delaney EK, Murchie P, Lee AJ, Ritchie LD, Campbell NC. Secondary prevention clinics for coronary heart disease: a 10–year follow-up of a randomized controlled trial in primary care. Heart. 2008;94(11):1419–23.

24 Denver EA, Barnard M, Woolfson RG, Earle KA. Management of uncontrolled hypertension in a nurse-led clinic compared with conventional care for patients with type 2 diabetes. Diabetes Care. 2003;26(8):2256–60.

25 Dierick-Van Daele ATM, Metsemakers JFM, Derckx EWCC, Spreeuwenberg C, Vrijhoef HJM. Nurse practitioners substituting for general practitioners: Randomized controlled trial. J Adv Nurs. 2009;65(2):391–401.

26 Flynn BC. The effectiveness of nurse clinicians' service delivery. Am J Public Health. 1974;64(6):604–11.

27 Hesselink AE, Penninx BW, van der Windt DA, van Duin BJ, de Vries P, Twisk JW, et al. Effectiveness of an education programme by a general practice assistant for asthma and COPD patients: results from a randomized controlled trial. Patient Educ Couns. 2004;55(1):121–8.

28 Houweling ST, Kleefstra N, van Hateren KJJ, Groenier KH, Meyboom-de Jong B, Bilo HJG. Can diabetes management be safely transferred to practice nurses in a primary care setting? A randomized controlled trial. J Clin Nurs. 2011;20(9/10):1264–72.

29 Jarman B, Hurwitz B, Cook A, Bajekal M, Lee A. Effects of community based nurses specialising in Parkinson’s disease on health outcome and costs: randomized controlled trial. BMJ. 2002;324(7345):1072–5.

30 Kinnersley P, Anderson E, Parry K, Clement J, Archard L, Turton P, et al. Randomised controlled trial of nurse practitioner versus general practitioner care for patients requesting “same day” consultations in primary care. BMJ. 2000;320(7241):1043–8.

31 Kuethe M, Vaessen-Verberne A, Mulder P, Bindels P, van Aalderen W. Paediatric asthma outpatient care by asthma nurse, paediatrician or general practitioner: Randomised controlled trial with two-year follow-up. Primary Care Respiratory Journal. 2011;20(1):84–91.

32 Lewis CE, Resnik BA. Nurse clinics and progressive ambulatory patient care. N Engl J Med. 1967;277(23):1236–41.

33 Murchie P, Campbell NC, Ritchie LD, Deans HG, Thain J. Effects of secondary prevention clinics on health status in patients with coronary heart disease: 4 year follow-up of a randomized trial in primary care. Fam Pract. 2004;21(5):567–74.

34 Murchie P, Campbell NC, Ritchie LD, Simpson JA, Thain J. Secondary prevention clinics for coronary heart disease: four year follow up of a randomized controlled trial in primary care. BMJ. 2003;326(7380):84.

35 Raftery JP, Yao GL, Murchie P, Campbell NC, Ritchie LD. Cost effectiveness of nurse led secondary prevention clinics for coronary heart disease in primary care: follow up of a randomized controlled trial. BMJ. 2005;330(7493):707.

36 Shum C, Humphreys A, Wheeler D, Cochrane MA, Skoda S, Clement S. Nurse management of patients with minor illnesses in general practice: multicentre, randomized controlled trial. BMJ. 2000;320(7241):1038–43.

37 Voogdt-Pruis HR, Beusmans GHMI, Gorgels APM, Kester ADM, Van Ree JW. Effectiveness of nurse-delivered cardiovascular risk management in primary care: A randomized trial. Br J Gen Pract. 2010;60(570):40–6.

38 Royal College of Nursing. Advanced Nurse Practitioners – an RCN Guide to the Advanced Nurse Practitioner Role, Competences and Programme Accreditation. In London: RCN; 2012.

39 Cooper R, Stoflet S. Diversity and consistency: the challenge of maintaining quality in a multidisciplinary workforce. J Health Serv Res Policy. 2004;9(Suppl 1):39–47.

40 Association of American Medical Colleges. Physician Shortages to Worsen Without Increases in Residency Training; 2011.

41 Cooper RA, Laud P, Dietrich CL. Current and projected workforce of non-physician clinicians. JAMA. 1998;280(9):788–94.

42 Delamaire M, Lafortune G. Nurses in Advanced Roles: A Description and Evaluation of Experiences in 12 Developed Countries, OECD Health Working Papers, 2010;No. 54, OECD Publishing. Available from http://dx.doi.org/10.1787/5kmbrcfms5g7–en

43 Aiken LH, Sloane DM, Bruyneel L, Van den Heede K, Griffiths P, Busse R, et al. Nurse staffing and education and hospital mortality in nine European countries: a retrospective observational study. Lancet. 2014, Feb 25. DOI: 10.1016/S0140–6736(13)62631–8.

44 Blegen MA, Goode CJ, Park SH, Vaughn T, Spetz J. Baccalaureate education in nursing and patient outcomes. J Nurs Adm. 2013;43:89–94.

45 Friese CR, Lake ET, Aiken LH, Silber JH, Sochalski J. Hospital nurse practice environments and outcomes for surgical oncology patients. Health Serv Res. 2008;43:1145–63.

46 Kutney-Lee A, Sloane DM, Aiken LH. An increase in the number of nurses with baccalaureate degrees is linked to lower rates of postsurgery mortality. Health Aff. (Millwood) 2013;32:579–86.

47 Tourangeau AE, Doran DM, McGillis Hall L, O’Brien Pallas L, Pringle D, Tu JV, Cranley LA. Impact of hospital nursing care on 30-day mortality for acute medical patients. J Adv Nurs. 2007;57:32–44.

48 Morgan S. What are the differences in nurse practitioner training and scope of practice in the US and UK? Nurs Times. 2010;106:21–4.

49 Leventhal ME, Denhaerynck K, Brunner-La Rocca HP, Burnand B, Conca-Zeller A, Bernasconi AT, et al. Swiss Interdisciplinary Management Programme for Heart Failure (SWIM-HF): a randomised controlled trial study of an outpatient inter-professional management programme for heart failure patients in Switzerland. Swiss Med Wkly. 2011;141:w13171.

50 Lugtenberg M, Burgers JS, Westert GP. Effects of evidence-based clinical practice guidelines on quality of care: a systematic review. Quality & safety in health care. 2009;18(5):385–92.

51 Thomas L, Cullum N, McColl E, Rousseau N, Soutter J, Steen N. Guidelines in professions allied to medicine. Cochrane Database Syst Rev. 2000(2):Cd000349.

52 Davis DA, Taylor-Vaisey A. Translating guidelines into practice. A systematic review of theoretic concepts, practical experience and research evidence in the adoption of clinical practice guidelines. CMAJ. 1997,157(4):408–16.

53 Sargent GM, Forrest LE, Parker RM. Nurse delivered lifestyle interventions in primary health care to treat chronic disease risk factors associated with obesity: a systematic review. Obes Rev. 2012;13(12):1148–71.

54 Rosemann T, Joest K, Korner T, Schaefert R, Heiderhoff M, Szecsenyi J. How can the practice nurse be more involved in the care of the chronically ill? The perspectives of GPs, patients and practice nurses. BMC Fam Pract. 2006;7:14.

55 Deeks JJ, Dinnes J, D’Amico R, Sowden AJ, Sakarovitch C, Song F, et al. Evaluating non-randomized intervention studies. Health technology assessment (Winchester, England). 2003;7(27):iii-x, 1–173.

56 Hoff TJ, Witt LC. Exploring the use of qualitative methods in published health services and management research. Med Care Res Rev. 2000;57(2):139–60.

57 Möhler R, Bartoszek G, Meyer G. Quality of reporting of complex healthcare interventions and applicability of the CReDECI list – a survey of publications indexed in PubMed. BMC Med Res Methodol. 2013;13(1):1–8.