Are community-based nurse-led self-management support interventions effective in chronic patients? Results of a systematic review and meta-analysis

The expansion of primary care and community-based service delivery systems is intended to meet emerging needs, reduce the costs of hospital-based ambulatory care and prevent avoidable hospital use by the provision of more appropriate care. Great emphasis has been placed on the role of self-management in the complex process of care of patient with long-term conditions. Several studies have determined that nurses, among the health professionals, are more recommended to promote health and deliver preventive programs within the primary care context. The aim of this systematic review and meta-analysis is to assess the efficacy of the nurse-led self-management support versus usual care evaluating patient outcomes in chronic care community programs. Systematic review was carried out in MEDLINE, CINAHL, Scopus and Web of Science including RCTs of nurse-led self-management support interventions performed to improve observer reported outcomes (OROs) and patients reported outcomes (PROs), with any method of communication exchange or education in a community setting on patients >18 years of age with a diagnosis of chronic diseases or multi-morbidity. Of the 7,279 papers initially retrieved, 29 met the inclusion criteria. Meta-analyses on systolic (SBP) and diastolic (DBP) blood pressure reduction (10 studies—3,881 patients) and HbA1c reduction (7 studies—2,669 patients) were carried-out. The pooled MD were: SBP -3.04 (95% CI -5.01—-1.06), DBP -1.42 (95% CI -1.42—-0.49) and HbA1c -0.15 (95% CI -0.32–0.01) in favor of the experimental groups. Meta-analyses of subgroups showed, among others, a statistically significant effect if the interventions were delivered to patients with diabetes (SBP) or CVD (DBP), if the nurses were specifically trained, if the studies had a sample size higher than 200 patients and if the allocation concealment was not clearly defined. Effects on other OROs and PROs as well as quality of life remain inconclusive.

Introduction education in a community setting on patients >18 years old with a diagnosis of chronic disease or multiple morbidity (see Table 1 for definitions). For this purpose, we drafted a protocol based on the population, intervention, comparison and outcome (PICO) approach [27] and the recommended guidelines for the reporting of systematic reviews and meta-analyses [28].
Studies aimed to evaluate the efficacy of a nurse-led self-management support intervention, compared to the usual care, to improve observer-reported outcomes (OROs) [29,30]-particularly clinical outcomes-and patient-reported outcomes (PROs) [30,31]as primary outcomes. We excluded studies that evaluated interventions in which nurses were only involved in Table 1. Definitions of setting and interventions.

Nurse-led Self-Management Support intervention
A systematic educational intervention that was targeted toward patients previously clinically assessed with a chronic disease.
Nurse assessed determinant to provide a tailored educational intervention through an holistic perspective, focused on preserving or enhancing health and patient's self-management goal achievement. Nurse provided health education to promote compliance and a healthy lifestyle. The intervention is finalized to help patient actively participate in either or both of the following: self-monitoring (of symptoms or of physiologic processes) or decision making (managing the disease treatment or exacerbation or its impact through self-monitoring).The intervention could be carried out by face to face encounters or consultation followed by telephone follow up. All telephone calls including prescriptions and patient concerns were addressed by the nurse who facilitated consultation with physician or other health professionals, if necessary.

Usual Care
Participants assigned to the usual medical care (control) group continued on-going care from their medical primary care provider (General Practitioner, Primary Care Physician) without any structured educational intervention.

Chronic diseases/ Noncommunicable diseases (NCDs)
Non-communicable diseases (NCDs), also known as chronic diseases, are not passed from person to person. They are of long duration and generally slow progression. The four main types of noncommunicable diseases are cardiovascular diseases (like heart attacks and stroke), cancers, chronic respiratory diseases (such as chronic obstructive pulmonary disease and asthma) and diabetes (http://www.who.int/mediacentre/factsheets/fs355/en/).

Nurses
Any qualified nurse working as a substitute to a primary care physician focused on Self-management support for chronic disease. This could include: nurse practitioners, clinical nurse specialists, advanced practice nurses, practice nurses, registered nurse, etc. As the job title, education, and experience of nurses varies considerably among and within countries. We did not select nurses by virtue of their job title but, based on the description of interventions and competencies (experience/training/ qualifications) we categorized nurses' roles into: (a) advanced practice nurse (APN) for example nurse specialist, nurse case manager and nurse practitioner and (b) registered nurse. We focused our interest mainly stressing the difference between basic and advanced level of nurse qualifications, to promote future comparison of job profile and a more efficient nurses insertion in the healthcare workforce.
medical assessment or therapy optimization and studies that enrolled patients with mental disorders. To ensure maximum retrieval, two reviewers with different skills in bibliographic search methodology and in nursing chronic disease management, searched together for RCTs in MEDLINE (to July 2016) using the strategy reported in S1 File. Additional searches in CINAHL, Scopus and Web of Science were carried out using similar syntax; experts were consulted and bibliographies of relevant articles were reviewed. Bibliographic search was restricted to studies reported in English. Each citation found in the databases was reviewed independently by two authors via a titles-first approach to obtain records for the abstract screening.

Study selection and quality assessment
Two reviewers independently reviewed the abstracts obtained in the search and retrieved the full text article of those that met the inclusion criteria. In cases of disagreement, full text article for review was retrieved. The methodological quality of the RCTs was assessed independently by two reviewers using the risk of bias approach described in the Cochrane Handbook [32]. Random sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting and other potential sources of bias were described and assessed. Any disagreements about methodological quality were resolved by discussion and, if necessary, a third reviewer was involved.

Data extraction
Two reviewers performed data extraction and data entry independently, in duplicate. Differences in data extraction were discussed and if necessary resolved by a third reviewer. A standardized form was used to abstract the following data: bibliographic details; population demographics; interventions; patient condition (diabetes, cardiovascular diseases (CVD), multichronic conditions); type of nurses employed in the study (RN: registered nurse; APN: advanced practice nurse); availability of specific training for the nurses that provide the intervention; type of intervention (face-to-face; telephone/telemedicine; mixed); duration of the intervention; study sample size; outcome data (continuous or binary).

Data synthesis
A rating system, based on the methodological quality of the studies and on the consistency of the findings [33,34], was used to assess the strength of the evidence for OROs and PROs. The results were synthesized and assigned one of the following three levels of scientific evidence: • strong evidence: provided by generally consistent findings, supporting the hypotheses, in multiple high-quality studies; • moderate evidence: provided by generally consistent findings, supporting the hypotheses, in one high quality study and one or more moderate quality studies, or in multiple moderate quality studies; • insufficient evidence: only one study available or inconsistent findings in multiple studies.
To summarize continuous data, the pooled mean difference (MD) and 95% confidence interval (CI) were calculated [35]. A random effect approach was chosen for all analyses to account for between-study variance [36]. The fixed-effects model [37] was also used to check the level of agreement with random effects conclusions. The I 2 metric, which describes the percentage of total variation across studies that was due to heterogeneity rather than sample error (chance) [38], was used to test for heterogeneity. If I 2 was !60%, a sensitivity analysis was performed by removing the studies contributing to the heterogeneity. Results of studies reported in multiple articles were included once in each meta-analysis. Presence of publication bias was assessed through funnel plot graph.
Given the highly diverse nature of the studies analysed, several stratified meta-analyses were carried out to explore the efficacy in subgroups; meta-analyses were also carried out in the absence of statistical heterogeneity. In particular, we analyzed the effect of the following stratification factors: patient condition (diabetes, cardiovascular diseases (CVD), multichronic conditions); type of nurses employed in the study (RN: registered nurse; APN: advanced practice nurse); availability of specific training for the nurses that provide the intervention; type of intervention (face-to-face; telephone/telemedicine; mixed); duration of the intervention ( 6 months; >6 months); study sample size ( 200; >200); attrition rate (<20%; !20%); allocation concealment (clearly stated; undefined/absent).
All meta-analyses were performed using RevMan software, version 5.2 (Cochrane Collaboration, Oxford, UK, 2012). Reporting was made following the PRISMA Statement guidelines (see S2 File for the Checklist).

Main characteristics of the included studies
Of the 7,279 papers initially retrieved (Fig 1) 29, that describe the results of 23 studies, met our inclusion criteria (see S1 Table for a summary of the main characteristics and an overall quality score of the studies included in the review). A summary of the type of intervention and primary outcomes measured in each study is reported in Table 2.
The studies were published from 2000 to 2013, mainly in the USA (15), the UK (5) and the Netherlands (4). Overall, 10,162 patients were enrolled in the 23 studies (range: 51-1665), seven of which enrolled fewer than 200 patients. Six papers [39][40][41][42][43][44] reported analyses of previous studies [45][46][47][48][49], which extended the follow-up and/or took into account different outcomes; these were included in the meta-analyses as appropriate.  (11), diabetes (9) or multichronic conditions (7). Only two papers took into account patients with COPD. Interventions were mainly carried out at patients' homes (10 studies) and in general practices (five studies) by APNs (13 studies) and RNs (10 studies); the nurses were specially trained in 15 studies. It is interesting to note that self-management skills were appropriately assessed in patients by validated tools in only five studies.
The methodological quality was high in nine studies and moderate in another nine (S2 Table). Only one paper fulfilled all the criteria for reducing risk of bias. Eight studies failed to report only one of the criteria. Nine papers out of 29 did not report on the methods used to randomly allocate patients to groups and in 20 and 11 cases the allocation concealment and the blinding, respectively, were not sufficiently detailed or were clearly absent. Five studies were at high risk of bias for attrition.
The majority of effective interventions were carried out by advanced nurses/case managers [43,48,52,53,55]. A variety of intervention techniques were used: four out of the seven effective studies used face-to-face studies [24,50,55] or face-to-face/telephone [53] nurse visits; these were delivered at the patient's home [50,53], in nurse-led clinics [24], at local community activity centres [53] or in primary care clinics [55].
Meta-analyses of subgroups showed a statistically significant effect if the interventions were delivered to diabetic patients (MD -2.56, 95% CI -4.82 to -0.31), if an APN was employed (MD -3.57, 95% CI -6.36 to -0.78), if the nurses were specially trained (MD -2.81, 95% CI -4.30 to -1.32), if the studies had a sample size greater than 200 patients (MD -0.13, 95% CI -0.25 to -0.01) and if the allocation concealment was not clearly defined (MD -2.54, 95% CI -5.04 to -0.56). Stratification by type of intervention failed to show a significant effect of any specific  Table 2. Summary characteristics of the intervention of included studies.

Bischoff et al., 2012, The Netherlands
The self-management program consisted of paper modules and a written exacerbation action plan. The practice nurse applied the program to the individual patient in two to four sessions of approximately one hour each, scheduled in four to six consecutive weeks, followed by telephone calls/General practice. Nurse-led self-management support interventions in chronic patients  (Table 3). Ten studies with 3,881 patients in total were included in the meta-analysis on the reduction in DBP [24, 39,44,48,[50][51][52][53][54][55]. A statistically significant reduction in DBP was found for the whole group (MD -1.42, 95% CI -1.42 to -0.49) with no statistically significant heterogeneity between studies (I2 = 34%, p = 0.14) (Fig 2). The analysis of the funnel plot showed a lack of studies with large sample size and high effect measures.
HbA1c. Of the 29 included studies, 11 [25, 39, 43, 48, 51, 54-59] investigated HbA1c levels as a primary outcome in diabetic patients, resulting in strong evidence of the efficacy of intervention. Overall, these studies included 4,207 patients. The levels of HbA1c were significantly lower in the experimental groups than in the control groups in four studies [25,43,48,58] ( Table 3). The two studies with statistically significant results and high methodological quality , involving 2,669 patients, were useful for pooling data. The MD showed a reduction in HbA1c of 0.15 in favour of the experimental group (95% CI -0.32 to 0.01) with a heterogeneity of I 2 = 28, p = 0.21 (Fig 2). The funnel plot showed that the results were based mainly on small studies with low-effect measures.
Total mortality. Three studies [45,60,61], with an overall sample size of 2,564 patients, evaluated total mortality. The study of Delaney et al. [41] used the same population and intervention as Murchie et al. [45] but considered the results from 10 years of follow-up. For all four studies the total number of deaths in the experimental groups was lower than in the control groups, reaching statistical significance in two studies [45,60] (Table 3); these studies were based on interventions lasting 12 months [45] or longer [60] on patients with coronary heart disease or chronic heart failure led by RNs [45] or APNs [60,61]. Educational interventions were based on face-to-face visits carried out at nurse-run clinics [45] or hospital [61] with telephone follow-up [60,61].
Multiple clinical outcomes. Only one study [62] evaluated as a primary outcome the simultaneous reaching of a threshold in BP levels, LDL serum levels and percentage of HbA1c, taking into account 556 patients. A significantly higher percentage of patients in the intervention group reached the goals compared to the control group. The intervention consisted of an initial personal meeting with a nurse case manager, followed by follow-up telephone calls. Nurse-led self-management support interventions in chronic patients

Patient reported outcomes
Quality of life. Three studies [40,63,64] included changes in quality of life-evaluated with SF-36 [40,64] or other questionnaires related to the specific disease aim of the study [63,64]-as a primary outcome, but there was insufficient evidence of a significant effect. The overall scores of the questionnaires were analyzed. For two studies [40,64] the overall scores in the experimental groups were higher rather than the control groups, but this result was only significant for the study of Murchie et al. [40] (Table 3). Educational interventions were based on Table 4. Meta-analysis of the reduction of blood pressure levels stratified by level and training of employed nurses; type and duration of the intervention; study size; attrition rate; allocation concealment. Nurse-led self-management support interventions in chronic patients face-to-face visits [40,63] or telephone health mentoring [64] led by RNs [40,64] or APNs [63].

Discussion and conclusions
Primary care systems across the world are facing the challenge of an ageing population and an associated increase in the number of chronic patients [65,66], leading to a growing demand for a kind of care [67] that meets emerging needs, reduces the costs of hospital-based ambulatory care and prevents avoidable hospital use by the provision of more appropriate care systems. In this context, the rational redistribution of tasks among health workforce teams-namely task shifting [15]-as a means of addressing this public health issue represents a potentially winning strategy. More particularly, serious attention has been payed to the support of patient self-management, since it can improve patient self-efficacy [8,68], disease-related behaviors and, finally, enhance patients' functional and health status [8,69,70]. Among health professionals, nurses can play a pivotal role in the delivery of self-management support interventions, particularly in areas of medical workforce shortage. This policy development clearly brings with it the need to continually seek updated evidence about the roles that nurses can undertake, their clinical effectiveness and cost-effectiveness in these roles, as well as patient satisfaction. According to our systematic review and meta-analysis, nurse-led self-management support interventions in chronic care community programs have a positive impact on some OROs, such as a reduction in the levels of HbA1c, DBP/SBP and, to a lesser extent, LDL, particularly in patients with diabetes and CVD. Effects on other outcomes such as serum levels of total cholesterol, fasting serum glucose levels and triglycerides, as well as quality of life and allcauses mortality, remain inconclusive. Diabetes and CVD are among the diseases that can most benefit from patient self-management. Empowering patients to manage their own diseases and fostering patient-centered activities can effectively reduce complications or reactivation of diseases that can shorten length of life and reduce autonomy. Self-management training in type 2 diabetes has evolved since the didactic primarily interventions of the 1970s into the empowerment models of the 1990s [69,71]. Such a transformation has led to better glycemic control [69]. Our results confirm this and suggest also that trained nurses can effectively administer self-management support interventions to type 2 diabetes patients [25,43,48,58]. A study published in 2004 showed that a nurse-led education intervention led to the improvement of glycemic control and a delay in the requirement for insulin therapy in patients treated with oral hypoglycemic therapy [58]. Moreover, our results show that nurse-led telemedicine interventions can also have a positive effect by reducing HbA1C levels [43,48]. The remote monitoring and transmission of physiological data facilitate contact with a health care professional via telephone or video, while disease-specific education guarantees the reinforcement of self-management behaviors [72]. More difficulties were encountered in reducing serum levels of LDL [25,39,43,48,55] and triglycerides [25] in patients with diabetes. This is of particular interest since LDL oxidation does not decrease after improvement in metabolic control in type 2 diabetes [73]. Together with hypertriglyceridemia, LDL oxidation is involved in the pathogenesis of the so-called metabolic syndrome, which is associated with increased risk of CVD and for which lifestyle modification is an important therapeutic strategy [74]. Therefore, developers of educational interventions should focus on general knowledge of diabetes, adherence to medication, lifestyle changes and, if possible, self-monitoring of blood glucose [75].
With respect to CVD, the results of our meta-analysis also show that nurses can be more effective than the usual care-in-the-community systems in improving blood pressure control, eventually leading to reduced blood pressure levels. This positive effect is clearer when nurses are specially trained and is more significant among diabetes patients for SBP levels and among CVD patients for DBP levels. Face-to-face interventions seem to be more effective, at least for the reduction of DBP levels, even though nurses also significantly improve self-management behavior by telephone interventions [47].
Nurse-led intervention is less effective at improving clinical outcomes in multi-chronic patients [24-26, 39, 44, 46, 49] probably because of the subjective and objective barriers to good self-management associated with this condition. Indeed, comorbidity has been mentioned in previous studies as a limit to self-care [76,77]. A semi-structured interview study concluded that major barriers to self-care for people with more than one chronic disease are mainly linked to the combined effects of multiple conditions or to a single dominant disease making the management of the other conditions difficult. Other barriers were identified as a lack of patient knowledge about their conditions, financial constraints, low self-efficacy, inadequate communication with providers, the need for or use of social support and finally various logistical issues [78]. Another qualitative study, which used patient focus groups, placed much more emphasis on the role of physician communication and family support as barriers to the self-management of their chronic conditions [79]. Clearly, self-care interventions for people with multiple chronic diseases must be tailored to patients' real needs, since they are likely to be more effective if targeted at particular risk factors or specific functional difficulties [80].
The finding that the benefits of nurse-led intervention to support patient self-management disappear when nurses are not specially trained is one of the most important results of this meta-analysis. Ad hoc training seems to be more important than the role and general experience of the nurse. In fact, the results of the meta-analyses show that APNs are more effective than RNs only in reducing SBP levels. Provider training is recognized to be a key factor in the entire self-management support intervention process. Studies that evaluated the effectiveness of in-person training have reported generally positive results [81][82][83]. However, promising results also derive from web-based self-management training for health professionals: webinar-based training sessions can benefit participants' self-beliefs and confidence [84].
Several studies have determined that, among health professionals, nurses are best placed to promote health and to deliver preventive programs within the primary care context [85,86]. Their employment as providers of self-management support programs in primary care can further improve the health status of chronic patients, even if the task shifting from physicians to nurses in this particular area requires specific education and training. Further research on the efficacy of nurse-led self-management support programs must focus on long-term outcomes. Evidence on the effect of these programs on mortality and hospitalization rates is still insufficient or lacking. Moreover, the evaluation of patient self-efficacy in experimental studies that use reliable and valid instruments is urgently required.
Finally, the methodological quality of RCTs must be improved. In many cases, in the particular context of trials that evaluate the efficacy of nurse-led interventions vs. physician-led interventions, blind participation in the intervention is not always possible. This was often acknowledged in the included studies, but it was not always counterbalanced by appropriate allocation concealment that, in such cases, is universally recognized to reduce bias [87].
Our systematic review and meta-analysis have several weaknesses that must be taken into account. First of all, we included only publications in English and we did not search for grey literature. However, we made the literature search as widespread and inclusive as possible; primarily, we used electronic databases, but also screened the bibliographies of the retrieved articles for relevant publications. Second, one may argue that some clinical and physiological characteristics of the patients other than the educational interventions could influence the outcomes. To reduce this possibility to a minimum, we included only RCTs because of their lower risk of bias and we used restrictive inclusion and exclusion criteria to minimize heterogeneity among patient populations in terms of severity of disease, learning abilities and capacity to realize autonomously the activities of daily living. However, future research that includes non-randomised trials and/or observational studies are strongly recommended. Finally, we included different types of intervention. We decided to use this strategy because even though self-management support interventions differ in terms of target population, mode, format and content, it is clear that this variability in approach does not markedly affect outcomes [88]. Moreover, we made stratified analyses to account for some characteristics of the interventions that might affect the results.
In conclusion, self-management is a key focus of health policies for chronic disease control in many countries. Nurse-led self-management support interventions can be included in routine primary care activities, since specially trained nurses appear to be more effective than physicians in educating patients with diabetes and CVD in self-management of blood pressure and Hb1Ac in community settings. Future research should evaluate the efficiency of task shifting from physicians to nurses in community settings. Furthermore, trials with higher methodological quality and larger patient populations are urgently needed to assess the efficacy of selfmanagement programs, since current evidence is based on very few large studies of mixed methodological quality.