Effectiveness of interventions involving nurses in secondary stroke prevention: A systematic review and meta-analysis

Eligibility criteria

We determined participants, interventions, comparator, and outcomes prior to initiating the study. The question for this study is: among adult patients diagnosed with stroke/TIA, do secondary prevention programs where nurses have a primary role affect risk factor modification? Included RCTs had secondary stroke prevention interventions where the role of a nurse was defined (e.g. supporting patients with individual goal-setting). Studies needed to have baseline and post-intervention results with continuous data to calculate standardized mean difference (SMD), or post-intervention categorical results to calculate odds ratios (ORs). We included interventions at any time after the stroke/TIA and analyzed the first set of outcomes collected after the end of the intervention (it was not common to have a second longer follow-up). The comparator was usual care, or in some studies, a control with fewer activities (e.g. a single lifestyle counseling session, educational materials).

The main outcome measures included in this review were: 1) medical risk factors, 2) behavioral risk factors, and 3) knowledge of risk factors among stroke survivors. Medical risk factors included both systolic (SBP) and diastolic blood pressure (DBP), HbA1c, lipids and cholesterol, while behavioral risk factors included dietary changes, physical activity, medication adherence, smoking cessation, and alcohol use. We excluded studies focusing only on the spouses or caregivers, studies comparing two types of interventions without a true control group, and studies where more than half of the study participants did not have a diagnosis of stroke.

Search strategies

We finalized a comprehensive literature search for English language articles published in peer-reviewed journals using major electronic databases (Medline, CINAHL, EMBASE, PsychINFO, Cochrane Database of Systematic Reviews, Web of Science, Proquest Dissertation) and including articles published until the end of 2017. Our Medline search strategy is summarized in Table 1. Some of the MeSH terms and keywords included “stroke”, “secondary prevention,” and “nursing.” The studies were screened for redundant articles by entering them into a computer-based reference management system (RefWorks). ¹⁰ A second reviewer (BMS) screened abstract titles and abstracts. Full texts of selected studies from the initial screening were obtained. When there were discrepancies, two reviewers (BPP and BMS) discussed the study to reach a decision. The reference lists of these articles were hand-searched for additional relevant references. We attempted to contact researchers to learn about unpublished trials to reduce publication bias and also to clarify additional data-related questions during analysis.

OPEN IN VIEWER

Table 1.

Search strategy.

Search steps	Searches	Results
1.	exp nursing care/	128477
2.	exp Nurses/	81882
3.	Nursing/ or rehabilitation nursing/	51,892
4.	(Nurse* or Nursing).mp. [mp=title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]	644,405
5.	1 or 2 or 3 or 4	644,956
6.	((cerebral or cerebrovascular or brain or vascular) adj3 (accident* or infarct*)).mp. [mp=title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]	54,170
7.	(stroke* or CVA)mp.	252,654
8.	exp stroke/	111,905
9.	6 or 7 or 8	285,223
10.	secondary prevention/ or tertiary prevention/	17,838
11.	((previous or prior or second* or subsequent or tertiary or recurrent*) adj3 (prevent* or avoid)).mp. [mp=title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]	59,232
12.	rehab*.mp. [mp=title, abstract, original title, name of substance word, subject heading word, keyword heading word, protocol supplementary concept word, rare disease supplementary concept word, unique identifier, synonyms]	168,108
13.	10 or 11 or 12	225,399
14.	5 and 9 and 13	1543
15.	randomized controlled trial.pt.	455,529
16.	controlled clinical trial.pt.	92,233
17.	randomized.ab.	405,256
18.	placebo.ab.	187,117
19.	drug therapy.fs.	1,999,990
20.	randomly.ab.	286,588
21.	trial.ab.	420,857
22.	groups.ab.	1,772,679
23.	15 or 16 or 17 or 18 or 19 or 20 or 21 or 22	4,158,427
24.	exp animals/ not humans.sh.	4,433,718
25.	23 not 24	3,593,245
26.	14 and 25	409
27.	Limit 26 to English language	383

Data extraction

Data from selected studies were extracted into a study specific data extraction table by two reviewers (BPP and BMS) (se Supplementary Material online). The Physiotherapy Evidence Database (PEDro) scale ⁹ was utilized to assess for the potential bias involved in the methodology of the studies. When available, we used the scores posted on the PEDro website. PEDro scores range from 0 to 10 and trials with scores ⩾ 6/10 on the PEDro scale are indicative of moderate to high quality trials. ⁹ When scores were not available, both reviewers assessed the study quality independently.

Statistical analysis

Studies reporting continuous data were meta-analyzed using the SMD. ¹¹ SMDs were pooled across studies for each outcome to obtain the pooled effect size. When medians and interquartile ranges were reported, the data were converted to means and standard deviations. ¹² We characterized the magnitude of the SMD as small (0.20), moderate (0.50) or large (0.80). ¹³ With dichotomous data, pooled ORs and 95% confidence interval (CI) were estimated using the Mantel–Haenzel method. We estimated the effects of continuous data by finding the differences between the pre-intervention and immediate post-intervention results. The I-squared (I²) value was used to quantify the statistical heterogeneity. ¹⁴ I² values of 25%, 50%, and 75% were considered low, moderate, and high heterogeneity, respectively. When statistical heterogeneity was 0% to 50%, the fixed-effects model was utilized. The random-effects model was used when I² values were over 50%. The p value was set at 0.05. Forest plots were created to graphically illustrate the results. All meta-analyses were conducted using RevMan 5. ¹¹

Sensitivity analysis

Sensitivity analyses were undertaken as part of assessing the robustness of the results. For this we excluded studies of lesser quality (less than 6 on the PEDro scale). ⁹

Search results and study description

Of 895 titles, 16 studies met inclusion criteria (Figure 1). Sixteen studies (14 independent trials) meeting inclusion criteria are included, with a total of 3568 stroke and TIA patients. Four works^15–18 reported findings based on two studies. Sample sizes ranged from 36 ¹⁹ to 533. ¹⁷ Below we present the quality of the evidence using the PEDro scores and summarize key elements from the Template for Intervention Description and Replication checklist to describe the interventions. ²⁰

OPEN IN VIEWER

Figure 1.

Flow diagram of the search and selection process.

Quality of evidence

PEDro scores ranged from 4 ²¹ to 8.^16,17,22,23 Five studies had low quality PEDro scores of less than 6.^{19,21,24–26}

Where the intervention took place

The studies were undertaken in the United Kingdom,^22,23 Sweden,^21,27 Denmark,^15,16 the USA,^19,28 Canada,^24,29 China,^25,30 Israel, ³¹ Australia,^17,18 and Taiwan. ²⁶ Most patients had just been discharged from hospital following their event, although some were three months ²² or nine months ²³ post-event. In seven studies^{15,17,19,21,27,28,30} researchers enrolled hospitalized patients, followed by interventions and outcome measurements in the community, while a few initiated studies from home, ²⁵ through stroke prevention clinic^{22–24,26,29} or rehabilitation centre. ³¹

Who provided the intervention

Interventions were conducted by stroke nurses,^21,29 senior nursing students, ³¹ stroke nurse specialist, ²² Advanced Practice Nurse, ²⁸ Clinical Nurse Specialists and Nurse Practitioners specialized in stroke and with special training in motivational interviewing, ²⁴ stroke trained nurses with training in motivational interviewing, ¹⁹ stroke nurses with no formal training in motivational interviewing,^27,30 and trained study nurses.^{15,17,23,25,26}

In six studies, nurses acted alone in delivering the interventions and therefore it was a new role for the nurses,^{15,22,23,25,26,30} although some worked in collaboration with physicians.^{17,19,21,27,28}

A few studies included other allied health disciplines, including a dietitian and an exercise physiologist, ¹⁹ physical therapist, ²⁸ nutritionist and social worker, ²⁹ physical therapist, occupational therapist, and social worker. ³¹

Most nursing interventions included specialized skills such as motivational interviewing and providing individual education on lifestyle changes including diet, physical activity, tobacco and alcohol use, the importance of medication compliance and its relevance to secondary prevention.^{15,17,19,21–31} In some studies, nurses supported patients with individual goal-setting and action planning to facilitate risk factor modification in alignment with chronic disease management principles.^{17,19,25,29,30} All studies focused on multiple components such as education, care planning, collaborating with physicians and other health team members, providing written and verbal information, medication review, arranging GP appointment, and sending discharge summary to the GPs.

Two studies^25,29 adapted stroke knowledge questionnaires from other studies and two studies^26,28 developed questionnaires for their study. We included questionnaires focused on the knowledge of risk factors for this meta-analysis, although one study ²⁹ combined knowledge of stroke symptoms, risk factors, and actions taken.

The modes of delivery of the intervention

The mode of delivery of interventions varied across the studies. Most studies focused on face to face meetings for delivering the interventions^{15,17,19,21,22,26,28,31} and a few had telephone based only,^27,30 both individual and telephone based interventions,^23,24 group sessions, ²⁵ and individual and group sessions. ²⁹ Three studies involved both patients and their caregivers in the interventions.^23,29,31

The control group received only usual care in eleven studies.^{15,17,19,21–23,25–27,30,31} Some studies provided some additional aspects, although they were not of the same magnitude as the intervention (e.g. single lifestyle counseling session; ¹⁶ educational materials; ²⁸ or physician assessment. ²⁴ ) No studies provided a true attention control where the amount of time and attention of the control group were matched with the intervention group.

Intensity of the interventions

The timing of the interventions ranged from 1–2 months ²⁵ to 12 months after discharge.^19,27 The duration of each intervention session ranged from 30 minutes to three hours.

Tailoring of the intervention

All studies followed a specific protocol in terms of interventions and follow-up assessments; however, care plan and the focus of the interventions were individualized to match the risk factor profile of the participants.

Adherence and fidelity

Most studies did not explain whether they assessed intervention adherence or fidelity. One multicenter study ²⁴ did describe the use of motivational interview scripts and investigator team meetings to facilitate consistency in recruitment and follow-up processes by the case managers.

Adverse events

In general, adverse events or deaths over the study were similar between groups. Six studies recorded vascular events and/or deaths and found similar events between groups.^{15,17,21,23,27,28} Four studies had small numbers of serious adverse events (1–2) in one group or the other.^19,24,29,30 Some vascular events or hospitalizations are expected when stroke patients are followed over several months given the age and multi-comorbidities that these patients have. One study ³¹ reported no hospital admission for either group over the study and another study ²⁵ reported that 25% and 19% had dropped out in the intervention group and control group respectively, citing hospitalization as one of the reasons for dropping out. Adverse event information was not reported in two studies.^22,26

Effect of interventions involving nurses on controlling medical risk factors

Meta-analysis of seven studies (n=1941)^{15,17,19,22–24,27} found no significant effect on SBP (SMD = −0.03 (95% CI = −0.26, 0.21), I²=82%; p=0.83) using a random-effects model (Figure 2). Pooled results from five studies (n=1372)^{15,22–24,27} demonstrated no significant effect on DBP using a random-effects model (SMD = 0.22 (95% CI = −0.20, 0.641); I²=93%; p=0.31) (Figure 2). However, one study ²⁴ was an outlier, with a large effect size favoring the control group for both SBP and DBP, while the other studies had small effect sizes favoring the experimental group. In addition, this study ²⁴ may have been weighted by outlier data; mean SBP and DBP increased by 7.2 and 22.2 mmHg, respectively, but median SBP and DBP decreased by 22 and 10 mmHg, respectively. After removing this study, the model produced a small but significant reduction of both SBP (SMD = −0.14 (95% CI = −0.23, –0.05), I²=0%; p=0.002) and DBP (SMD = −0.16 (95% CI = − 0.27, –0.05), I²=0%; p=0.003) using a fixed-effects model.

OPEN IN VIEWER

Figure 2.

Medical risk factors: effect sizes, standardized mean differences and forest plots.

Effect of nursing interventions on controlling behavioral risk factors

Three studies (n=425) using continuous data showed that the interventions had a small, significant effect on improving diet in a fixed-effects model (SMD = −0.21 (95% CI = −0.40, –0.02), I²=33%; p=0.03)^25,30,31 (Figure 3). Five studies (n=1233) reported on the number of physically inactive people post-intervention.^{17,19,25,28,29} A pooled random-effects model found that the intervention reduced the proportion of physically inactive people (OR = 0.60 (95% CI= 0.37, 0.97), I²=56%; p=0.04). Two studies (n=270) reported that the interventions had a significant effect on improving medication adherence in a fixed-effects model (SMD=0.41 (95% CI=0.17, 0.65), I²=0%; p=0.0009).^25,30

OPEN IN VIEWER

Figure 3.

Knowledge and behavioral risk factors: effect sizes, standardized mean differences, odds ratios and forest plots.

The role of nurses in smoking cessation through education and counseling was explored in six studies.^{16,17,19,21,28,29} A fixed-effects model including these six studies (n=1592) found no significant effect on smoking cessation (OR = 1.12 (95% CI= 0.87, 1.45), I²=0%; p=0.36). The role of nurses in reducing the use of alcohol was explored in three studies and found no significant effect (OR= 0.86 (95% CI= 0.46, 1.60), I²=0%; p= 0.63).

Effect of nursing interventions on improving knowledge of risk factors of stroke

A random-effects model of three studies^25,26,29 (n=516) on the role of nurses in improving the knowledge of stroke risk factors found a significant effect (SMD = 0.73 (95% CI= 0.28, 1.18), I²=83%; p=0.001) (Figure 3).

Sensitivity analyses

Sensitivity analysis was undertaken by eliminating two studies^19,24 that scored less than 6 on the PEDro scale for blood pressure outcomes. This resulted in significant fixed-effect models for SBP (SMD = −0.13 (95% CI −0.22, −0.04), I²=0%; p=0.004) and DBP (SMD = −0.16 (95% CI −0.27, −0.05), I²=0%; p=0.003). However, these results are likely a result of removing one study ²⁴ which was an outlier with a large effect supporting the control group. After excluding two studies of lower methodological quality^19,21 for the outcome of smoking cessation, the model (four studies, n=1165) still did not demonstrate a significant result (OR = 1.15 (95% CI 0.86, 1.54), I²= 0%; p=0.35).

Effect of nursing interventions on controlling other risk factors

SMDs and OR for other outcomes with two studies are shown in Figures 2 and 3 and are considered less robust results due to the small number of studies. Costs were not assessed in any of the studies.

Limitations

Methodological differences, which highlight the breadth of possibilities in which nursing interventions may be implemented, limit comparisons between studies in this review. There is variation in study design, interventions, mode of delivery, frequency and duration of intervention and follow–up, sample sizes, instruments used, involvement of caregivers, and variability in the members of the team and training and experience of the nurses. Individual studies may be biased by a lack of an attentional control group. The heterogeneity of the study populations, which include those with TIA, ischemic stroke and/or hemorrhagic stroke with a spectrum of functional impairments and in different phases of recovery, may further impact our study’s conclusions.