Electronic searches

The protocol stated that the Current Contents ISI and Web of Science ISI would be searched, but we did not search these for the review. The review authors believe that it would not add further studies to the identified titles.

Authorship

The protocol stated that the decision to carry out meta‐analyses were to be made by consensus of RC and JW. However, an additional author (LM) joined the review team at the review stage. Therefore, the decision to carry out meta‐analyses were made by consensus of RC, JW and LM.

Measures of treatment effect

For individual trials, for dichotomous (binary) outcomes, we planned to report odds ratios (ORs) and 95% confidence intervals (CIs). For continuous outcomes with different scales of measurement across trials, we planned to report standardised mean differences (SMDs), each with its 95% CI. In this review, we did not report on any dichotomous outcomes or continuous outcomes with different scales of measurement across trials.

Data synthesis

If there were sufficient, appropriate studies, they were to have been categorised based on study design, type of intervention or whether the intervention was aimed at patients or carers. Within these categories, the results were to have been further structured to reflect the comparisons detailed in the Types of interventions sections (i.e. mode of delivery). We also planned to present separately the results of studies that compared the intervention to no intervention, then those that compared the intervention to other forms of orientation intervention (e.g. face to face versus audio/visual) and those that compared two or more types of mode (e.g. written materials and video; written material and face to face). However, we did not carry out these procedures due to insufficient studies/data.

If cluster randomised trials were included, we would have accounted for the effects of clustering by adjusting each trial to its 'effective sample size’ using intra‐class coefficients where available, or external estimates from similar studies. However, there were no cluster randomised trials identified in this review.

Unit of analysis issues  

There were no unit of analysis issues.

Sensitivity analysis

We planned to restrict the primary analysis to studies which were considered as having a low risk of bias (i.e. those receiving a 'Yes' rating for the criteria of sequence generation and allocation concealment).

We planned to perform sensitivity analyses where appropriate in order to explore the influence of the following factors on effect size:
• excluding unpublished studies;
• excluding any large studies to establish how they impact on the results;
• excluding studies using the following filters: criteria used for clinical diagnosis and eligibility for intervention,
language of publication, country;
• the length of the interval between registration to the service and delivery of the intervention; and between delivery of the intervention and measurement of the effect.

We also planned to test the robustness of the results by repeating the analysis using different measures of effect size (risk difference, odds ratio etc.) and different statistical models (fixed‐effect and random‐effects models), as appropriate.

However, there were too few studies to perform these analyses.

Dealing with missing data  

If some outcome data remained missing despite our attempts to obtain complete outcome data from authors, we would have performed an available‐case analysis, based on the numbers of patients for whom outcome data were known. If standard deviations were missing, we would have imputed them from other studies, or where possible, computed them from standard errors using the formula SD=SE x √‾N, where these were available (Higgins 2008). We also planned to report on levels of drop outs in the intervention and comparison groups as an indicator of 'acceptability' of the intervention, and as a potential source of bias.

Assessment of reporting biases  

Reporting bias was to have been assessed using guidelines in Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). However, there were not enough studies available to do a meaningful assessment of publication bias.