Health professionals

We will include studies that describe care provided by health professionals providing direct services to patients admitted with acute stroke, working within acute stroke units (see Types of settings). Included participants will be licensed, registered healthcare providers, such as nurses, physicians, pharmacists, physiotherapists, occupational therapists, speech pathologists, dietitians, social workers, psychologists and radiographers.

Patients

We will include studies assessing care provided in acute stroke units within the first seven days poststroke onset, in all types of ischaemic and haemorrhagic strokes. We will include studies evaluating care in participants drawn from mixed diagnostic groups including stroke, but only where it is clearly possible to extract the separate results for people with stroke.

Types of settings

Eligible acute stroke units will be those that admit patients with acute stroke (usually within hours of onset), and where a multidisciplinary team, including specialist nursing staff based in a discrete ward, cares exclusively for stroke patients. We will use the Stroke Unit Trialists' Collaboration 2013 hierarchy of stroke service organisation to define and compare 'more‐organised' versus 'less‐organised' acute stroke services. The hierarchy in descending order from most organised is as follows (Stroke Unit Trialists' Collaboration 2013).

1. Acute stroke units that accept patients acutely but discharge early (usually within 7 days); these appear to fall into three broad subcategories:

   1. 'intensive' model of care, with continuous monitoring, high nurse staffing levels, and the potential for life support;

   2. 'semi‐intensive', with continuous monitoring, high nurse staffing but no life support facilities; and

   3. 'non‐intensive', with none of the above.

2. Comprehensive (i.e. combined acute and rehabilitation) stroke units that accept patients acutely, but also provide rehabilitation for at least several weeks if necessary.

To differentiate our review from the ongoing Cochrane Review by Cahill (in press) on evidence implementation in stroke rehabilitation settings, we will exclude settings described by Stroke Unit Trialists' Collaboration 2013 as rehabilitation stroke units that accept patients after a delay, usually of seven days or more, and that focus on rehabilitation.

Where insufficient detail is available in publications to determine the type of setting, we will contact authors.

Primary outcomes

1. Quality of care (as measured by health professionals' performance in terms of adherence to recommended practice or process of care). For example, the uptake or increase in:

   1. recommended diagnostic procedures or assessments;

   2. acute medical interventions;

   3. interventions to prevent complications;

   4. patient‐centred goal setting;

   5. early rehabilitation interventions;

   6. prescribing patterns for secondary prevention medications;

   7. assessments for post‐acute rehabilitation;

   8. referral patterns within the acute setting or to downstream services.

Secondary outcomes

Where reported we will consider summarising information on the following.

1. Patient outcomes, including mortality, morbidity, disability levels, medical complications, quality of life, or health benefit measures used in economic analyses such as quality‐adjusted life years.

2. Utilisation, coverage or access outcomes, for example, length of stay.

3. Resource use or economic outcomes.

   1. Direct medical costs.

   2. Non‐direct medical costs, such as out‐of‐pocket expenses.

   3. Indirect costs, such as productivity impacts from inability to work.

   4. Incremental cost‐effectiveness, cost‐utility, or cost‐benefit impacts of an intervention versus the comparator.

4. Health professional knowledge.

   1. Attitudes toward, or intentions to use evidence‐informed recommendations.

   2. Change in the knowledge or attitudes of acute stroke unit professionals regarding recommended acute stroke management.

We will include studies with measures of these secondary outcomes in this review. However, we will exclude studies that only report these outcomes with no objective measure of professional behaviour.

Databases

• Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library, Wiley (current issue)

• MEDLINE and MEDLINE In‐Process and other non‐indexed citations, OvidSP (1950 onwards)

• Embase OvidSP (1947 onwards)

• CINAHL (Cumulative Index to Nursing and Allied Health Literature), EbscoHost (1980 onwards)

• The Joanna Briggs Institute EBP Database, OVID SP (1998 onwards)

• ProQuest Dissertations & Theses Full Text (all dates)

The OVID MEDLINE search strategy provided in Appendix 1 is a draft developed by one author in consultation with a research librarian at the University of South Australia. We will present the finalised strategy in the review. We will adapt the MEDLINE strategy for other databases using appropriate syntax and vocabulary for those databases. We will not limit the search by date or language.

Grey literature

We will conduct a grey literature search to identify studies not indexed in the databases listed above. Sources will include the sites listed below. We will document additional sources, if any, in the review.

• Open Grey (www.greynet.org/opengreyrepository.html).

• Grey Literature Report (New York Academy of Medicine) (www.greylit.org).

• Agency for Healthcare Research and Quality (AHRQ) (www.ahrq.gov).

• National Institute for Health and Clinical Excellence (NICE) (www.nice.org.uk).

• Stroke associations/foundations websites.

Trial Registries

We will search the following registries for ongoing and completed trials.

• International Clinical Trials Registry Platform (ICTRP), World Health Organization (WHO) (www.who.int/ictrp/en).

• ClinicalTrials.gov, US National Institutes of Health (NIH) (clinicaltrials.gov).

We will also:

• review reference lists of all included studies, relevant systematic reviews and primary studies;

• contact corresponding authors of relevant studies or reviews to assist with identification of unpublished or ongoing studies; and

• conduct forward citation searching of included studies.

Contacting corresponding authors

Where insufficient published data are available, one review author will contact the study's corresponding author (for example, to determine details of the types of settings, the components and complexity of the implementation intervention, or to request further information or access to unpublished results).

Scope of the implementation intervention

We will extract data to describe the scope of the intervention, specifically whether the intervention was targeted at a single acute stroke unit at one study site or at single acute stroke units at multiple study sites with an interorganisation intervention component (such as a regional stroke management improvement collaborative).

Organisational context framing the intervention

We will extract descriptions of the healthcare setting(s), including information to decide the type of acute stroke unit as per the Stroke Unit Trialists' Collaboration 2013 categories (acute intensive, acute semi‐intensive, comprehensive or integrated traditional Chinese medicine stroke units). In our analysis, we will include type of acute stroke unit as a potential effect modifier (Stroke Unit Trialists' Collaboration 2013). Other descriptive data extracted will include size of unit (number of patients with stroke admitted per year, number of beds allocated to stroke), urban or rural setting, public/private health insurance funding, and level of advantage or disadvantage (e.g. socioeconomic characteristics of the setting).

Components and complexity of the implementation intervention

We will extract data on the intervention components using a framework based on the Cochrane EPOC taxonomy to guide data extraction (EPOC 2016a). Data will include specific tools or processes used in the implementation, categorised as either delivery arrangements, financial arrangements, governance arrangements or implementation strategies.

We will extract descriptions of the intervention(s) and implementation methods used to improve health professionals' performance (planned and actual), and classify studies as single intervention strategies or multifaceted intervention strategies (two or more interventions). This is because we are interested in contributing to building the understanding of whether there is a differential effect between single and multifaceted interventions (Squires 2014).

We will also extract data on the intervention duration, the number and composition of participating acute stroke professionals including professional disciplines, cost of the implementation intervention (where reported), and details of the implementation intervention including content, personnel delivering the intervention, delivery method and duration.

Complexity of the targeted professional performance change

For each study, we will report on the stated purpose of the targeted change (e.g. appropriate performance based on evidence‐informed clinical practice guidelines, cost containment) and the nature of the desired change (e.g. reduction, increase, cessation). Two review authors will independently categorise the complexity of the targeted change in a subjective manner as high, moderate or low using the method proposed by Brennan 2009. We will resolve disagreements by discussion among all review authors. The categories will be based on the:

• number of changes required;

• extent to which complex judgements or skills are necessary;

• number of staff and professions involved in the change; and

• number of facilities or departments involved in the change.

Outcomes

Outcome categories may include dichotomous or continuous measures of health professional performance; health professionals' knowledge, attitudes or intentions; or patient or system outcomes. For each category, we will include the primary outcome as identified by the trial authors. Where possible for randomised trials, we will verify that the primary outcomes reported in the publications are consistent with those specified in the trial protocols or published trial registration information. When trial authors do not specify primary outcomes, we will use the one specified in the sample size calculation. If there are no primary outcomes specified and no sample size calculations, or if there are multiple primary outcomes reported, we will rank the intervention effect estimates as reported in the publication, selecting the outcome that provides the median value for the effect estimate. When there is an even number of outcomes, we will include the outcome whose effect estimate is ranked n/2, where n is the number of outcomes. In the results tables, we will report whether we have used the authors' nominated primary outcome or the outcome providing the median effect estimate.

We will collect and report all outcomes described by trial authors in results tables, along with how they were measured (for example self‐report, chart audit).

Measures of treatment effect for randomised trials, cluster‐randomised trials, non‐randomised trials and controlled before‐after studies

We will extract the intervention effect estimate for included outcomes reported in the publications along with its P value and confidence interval (CI) or interquartile range, as appropriate, and the statistical analysis method used to calculate these measures. If trial authors used an inappropriate statistical method, we will not present the P value or confidence interval except if we are able to re‐analyse the data. In this circumstance, the P value will be annotated with the word 're‐analysed' in the results tables.

To make comparisons between studies, where possible, we will calculate the effect estimates. For binary outcomes, our primary effect estimate will be the risk ratio (RR), and for continuous outcomes our primary effect estimate will be the standardised mean difference (SMD). We will calculate P values for these effect estimates, adjusting appropriately for the design, where possible. We will standardise the effect estimates so that ratios greater than one, and differences between the intervention and comparator groups greater than zero, represent benefit for the intervention group (Brennan 2009).

If we cannot summarise results as above, we will report them as part of a narrative of 'other data', and we will not include them in a meta‐analysis (Deeks 2011).

Measures of treatment effect for interrupted time series studies

For interrupted time series studies, we plan to report the following estimates and their P values from regression analyses that adjust for autocorrelation: change in level of the outcome at the first point after the introduction of the intervention (immediate effect of the intervention); and the postintervention slope minus the pre‐intervention slope (long‐term effect of the intervention).

Clustering

For studies where clusters of individuals are randomised (cluster‐randomised trials, controlled clinical trials) or allocated (controlled before‐after trials) to intervention groups, but where inference is intended at the level of the individual, the analysis will need to account for correlation of observations within clusters (Brennan 2009). The use of standard statistical methods assumes independence of observations and in clustered studies can result in artificially small P values and overly narrow CIs for the effect estimates (Ukoumunne 1999). We will seek assistance from a statistician where trial authors have used inappropriate statistical methods to assist us with re‐analysis of the data. If re‐analysis is not possible, we will report the effect estimate and annotate the phrase 'unit of analysis error'.

Interrupted time series studies

We will consult a statistician to assist in re‐analysis of results from interrupted time series studies whose authors used incorrect statistical methods that did not account for the autocorrection of data points (Ramsay 2003). We will obtain data from graphs or tables within the publication, or from the trial authors, where possible. We will use either regression analysis with time trends before and after the intervention, adjusted for autocorrelation and any periodic changes, or autoregressive integrated moving average (ARIMA) analysis. We will present the results for the outcomes as changes along two dimensions: change in level and change in slope. Change in level is the immediate effect of the intervention and is measured as the difference between the fitted value for the first post‐intervention data point (e.g. one month after the intervention) minus the predicted outcome one month after the intervention based on the pre‐intervention slope only. Change in slope is the change in the trend from pre‐ to postintervention, reflecting the 'long‐term' effect of the intervention (EPOC 2013b; Ramsay 2003).

'Summary of findings' table and assessing the certainty of the evidence

We will use the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness, and publication bias) to make judgements on the certainty of the available evidence (high‐certainty, moderate‐certainty, low‐certainty, and very low‐certainty) for each main outcome (Guyatt 2011). Two review authors will independently carry out this assessment, resolving discrepancies by discussion, or with a third review author as arbiter if required. We will present the information in a 'Summary of findings' table along with key information on the findings for each outcome including RR, comparative risks and the number of participants (Higgins 2011). We will justify all decisions to down‐ or upgrade the certainty of the evidence in relation to each outcome within footnotes.

We will use the GRADE software to generate the 'Summary of findings' table and the EPOC worksheets (EPOC 2013c; GRADEpro GDT 2014). The 'Summary of findings' table will display evidence for the primary outcome domain (quality of care) and four secondary outcome domains (patient outcomes; utilisation, coverage or access outcomes; resource use or economic outcomes; and health professionals' knowledge and attitudes.) If during the review process, we become aware of an important outcome that we failed to list in our planned 'Summary of findings' table(s), we will include the relevant outcome and explain the reasons for this in the section 'Differences between protocol and review'. As we cannot anticipate which specific outcomes our included studies will use, we will attempt to include at least one outcome from each of these four domains. As well as presenting the findings by outcome, we will also present the findings by study design in the 'Summary of findings' table.

We will consider whether there is any additional outcome information that was not able to be incorporated into meta‐analyses and note this in the comments and state if it supports or contradicts the information from the meta‐analyses. If it is not possible to meta‐analyse the data we will summarise the results in the text.