Types of studies

We will include randomised controlled trials, quasi‐randomised controlled trials, controlled before‐and‐after (CBA) studies and interrupted time series (ITS) studies.

CBAs will be included if they have at least two intervention sites and two appropriate control sites, and contemporaneous data collection. ITS will be included if they have a clear time point when the intervention occurred and at least three data points before and three after the intervention is introduced. We anticipate the inclusion of cluster RCTs as interventions to support palliative care volunteers are commonly group‐based.

We will also include any qualitative studies conducted concurrently with included trials for additional narrative analysis related to the acceptability of the intervention and issues related to its feasibility.

Types of participants

We will include studies that involve participants either as volunteers through a formal volunteer program in services offering palliative care in people's homes, residential aged care facilities or other settings in the community; or studies that involve palliative care clients or their families receiving the services of such volunteers not in hospital.

We will exclude studies involving informal carers and other volunteers not associated with a formal program. We will also exclude studies involving people described as lay health workers or trained health professionals working without pay.

Types of interventions

We will include all formal training and support programs for palliative care volunteers. We will classify programs or strategies in included studies according to any stated or implied purpose; that is, whether they are intended to build skills for the volunteer’s role, to enhance their coping, or to maintain service standards.

We will classify interventions to train or support palliative care volunteers in included studies according to any stated or implied purpose that addresses skills, coping or service quality. If necessary, we will give programs that do not fit these classifications a post hoc categorisation.

Training to develop skills may include active listening, advance care planning, or providing bereavement support. Strategies to support coping could include screening for specific roles, mentorship, or linking with volunteer networks. Strategies to maintain service quality may introduce codes of conduct or raise awareness of ethical issues.

We will consider orientation programs that provide information about the palliative care service, but do not aim to build capacity in the volunteers in terms of skills, coping capacity or with a specific purpose of maintaining service quality to be 'usual care'. We will include any comparison group in included studies.

Types of outcome measures

Electronic searches

We will search the following electronic databases:

• The Cochrane Consumers and Communications Review Group Specialised Register;

• The Cochrane Specialised Register of Controlled Trials (CENTRAL, The Cochrane Library, Latest Issue);

• MEDLINE (OvidSP) (inception to present);

• EMBASE (OvidSP) (inception to present);

• PsycINFO (OvidSP) (inception to present);

• CINAHL (EbscoHOST) (inception to present);

• ProQuest Dissertations and Theses (inception to present).

We will also search the Database of Abstracts of Reviews of Effects (DARE, The Cochrane Library) to identify potentially relevant reviews.

We will use the strategy in Appendix 1 to search MEDLINE, and will adapt the strategy for other databases, using the most appropriate controlled vocabulary for each. The strategy combines two components: effects of programs to support needs of palliative care volunteers, and impact on clients, families, health services and volunteers.

Searching other resources

For all included studies, we will search reference lists and conduct author and citation searches in Science Citation Index database.

Selection of studies

One review author (DH) will screen all titles identified by the search to exclude clearly irrelevant references. All other review authors will independently screen a randomly selected subset of studies to ensure consistent application of inclusion criteria. Any discrepancies will be discussed among authors and inclusion will be based on consensus.

At least two review authors (DH and AS) will independently screen all remaining titles and abstracts to identify potentially relevant papers. Full text copies of all potentially relevant papers will be retrieved, including those where description of the intervention, study design, participants, or outcomes is insufficient for a decision about inclusion. One review author will collate retrieved papers (AS) and two review authors will independently apply inclusion criteria (DH and LP). Disagreement will be resolved by discussion between all review authors.

Data extraction and management

It is anticipated that the heterogenous nature of the studies in this area will limit the opportunity for meta‐analysis and it is likely a narrative synthesis will be necessary.

At least two review authors (DH and LP) will extract data independently for each included study using a data collection checklist based on the Cochrane Consumers and Communication Data Extraction template, which will be pre‐determined and piloted by all review authors. Discrepancies in data extraction will be resolved by discussion between all review authors.

Assessment of risk of bias in included studies

Two review authors (DH, LP) will assess the risk of bias for each included study independently, using the ‘Risk of bias’ tool described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

We will assess risk of bias associated with RCTs. quasi‐RCTs and CBAs for seven domains from the ‘Risk of bias’ tool: random sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective reporting; and other sources of bias (Higgins 2011). For clustered studies, we will assess the risk of bias associated with an additional domain: selective recruitment of participants.

We will assess risk of bias associated with ITS studies for the following seven domains: intervention independent of other changes; shape of intervention effect pre‐specified; intervention unlikely to affect data collection; blinding of outcome assessors to intervention allocation; incomplete outcome data; selective outcome reporting; and other sources of bias (EPOC 2009).

We will judge whether a study has a low, high, or unclear risk of bias for each domain using the criteria specified in the Cochrane Handbook for Systematic Reviews of Interventions and Cochrane EPOC Group guidance. We will report our assessment of risk of bias for each domain for each included study with a descriptive summary our judgment.

Measures of treatment effect

Continous outcomes

We will extract or generate the mean difference (MD) from the means and standard deviations for continuous outcomes. These include:

• Three patient outcome categories: evaluations of care; psychological outcomes; and pain‐related outcomes.

• Two volunteer outcomes: evaluations of support; and psychological outcomes

• Three palliative care service outcomes: cost; volunteer recruitment, adverse events.

Dichotomous outcomes

We will extract or generate odds ratios (ORs) and their 95% confidence intervals (CIs) for dichotomous outcomes. Two outcome categories will be treated as dichotomous outcomes. Categorical data will be collapsed:

• The patient outcome category, care outcomes: for example place of death will be [home, not at home] or [consistent with preferences, not consistent with preferences] where more than two places of death are recorded.

• The volunteer outcome category: skill acquisition. Any categorical data will be collapsed [Skills acquired, skills partly or not acquired].

We will adopt the approach to the selection of measures of treatment effect as outlined by Brennan and colleagues (Brennan 2009) and include any primary outcome identified by study authors that falls within each of our outcome categories. We will also report any common outcomes within these categories, but these results will be additional findings.

If multiple primary outcomes are identified within any category, we will rank the reported effect estimates for each of these outcomes and select the outcome with the median effect estimate.

If no primary outcome within our categories is specified, we will adopt the following strategy. First we will use any outcome within our categories specified in sample size calculations; then, if necessary, we will rank relevant intervention effect estimates, as reported, and select the median effect estimate. If the number of outcomes is even (n), we will include the outcome whose effect estimate is ranked n/2.

We will report whether we have used the primary outcome or the outcome with the median effect estimate in the results tables. We will also verify whether the specified primary outcomes in included studies are consistent with those identified in trial protocols and/or trial registry entries. This will be possible for RCTs only.

Unit of analysis issues

Dealing with missing data

We do not plan to undertake any imputation for missing outcome data but we will impute missing summary data where possible and report any assumptions in the results tables. We will investigate the effect of our choice of ICCs on the pooled effect estimate in any meta‐analysis through sensitivity analyses.

Assessment of heterogeneity

If there is substantial heterogeneity in the types of training or support for palliative care volunteers across included studies, we will investigate it using subgroup analyses and sensitivity analyses. We will consider whether an overall summary is meaningful, and if it is, use random‐effects analysis to produce it. Explanation of these assessments will be recorded. We will assess statistical heterogeneity in any meta‐analyses by visually inspecting the scatter of effect estimates on forest plots and by the calculated I² statistic (Higgins 2003).

Assessment of reporting biases

In addition to an extensive search of the published and grey literature, we will search the International Clinical Trials Registry Platform Search Portal and Current Controlled Trials metaRegister of Controlled Trials to identify other trials that may not be published or are currently underway and contact the relevant investigators for further information. As stated above we will include all retrieved trials from this search in the review, and note those for which we cannot get further information. If we find sufficient RCTs (at least ten) we will use funnel plots and formal statistical tests to investigate funnel plot asymmetry. We anticipate heterogeneity in effect estimates because of variability in the interventions. We plan to use the approach identified by Brennan 2009, and developed by Rücker and colleagues, which proposes a statistical test for when substantial between‐study heterogeneity exists when data points are proportions (Rücker 2008).

Data synthesis

We will report summary statistics for each comparison from the included studies (RCTs, quasi‐RCTs, CBAs) including: baseline and follow‐up summary statistics; effect estimates and their statistical significance; and information on effect modifiers and study design. We will summarise effect estimates for dichotomous outcomes within each comparison and study design. This will include the median effect estimate, inter‐quartile range, and range. We will use graphs to display data graphically.

If possible, we plan to use meta‐analytical methods to pool relative risks measuring the effectiveness of different strategies to support palliative care volunteers focused on skill development and/or coping enhancement and/or service quality using orientation programs as the comparator. Random‐effects meta‐analysis will be used to pool intervention effects because of anticipated clinical and methodological diversity. We plan to report an approximate 95% range of underlying effect estimates, based on the between‐study variance estimate, to provide some information on the spread of effect estimates (Higgins 2009). We will prepare a 'Summary of findings' table from the results of the meta‐analysis using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2008).

We will present results from ITS studies in tables for each comparison, with summary statistics for each of the included studies, change in level of the outcome at the first point after the introduction of the intervention, post‐intervention slope minus the pre‐intervention slope, and information on effect modifiers. This will also be presented graphically using, for example, scatter plots of change in level versus change in slope with combinations of statistical significance denoted by different symbols.

Subgroup analysis and investigation of heterogeneity

We plan to carry out sub‐group analyses based on differences in the focus of interventions (skill development, coping enhancement or service standards) and to compare single purpose programs with multipurpose programs (for example, we will compare training for skill development only with programs that combine skill development and coping enhancement and/or service standard issues). If there are sufficient trials we will use random‐effects meta‐regression to estimate the relative change in the intervention effect (and 95% CI) for each.

Sensitivity analysis

We plan to undertake sensitivity analyses based on the risk of bias assessment. If possible studies at the greatest risk of bias, as described in the risk of bias tables, will be removed from the analysis.