Understanding the contribution of intervention components: a network meta-analysis approach to psychological preparation for surgery

  • R. Powell
  • S. Freeman
  • N. Scott
  • A. Sutton
  • N. Cooper
  • A. Manyande
  • C. Vögele
  • J. Bruce
  • L. Byrne-Davis
  • M. Johnston


Background: Our 2016 Cochrane systematic review and standard meta-analysis of randomised controlled trials found that patients receiving psychological preparation had reduced length of stay (LOS) but could not unpick the impact of different intervention components (e.g. procedural information, sensory information, relaxation, behavioural instruction). The present study used a network meta-analysis approach to explore which components, and combinations of components, were most effective. Methods: Component-based network meta-analyses using WinBUGS following Welton et al. (2009) examined four models: 1) Single treatment effect: all interventions were grouped together and compared with control. 2) Additive main effects: assumed each intervention component has a separate effect. 3) Two-way interaction: model 2 with additional terms for combinations of components. 4) Full interaction model: each combination of interventions was considered. Findings: Model 1 identified a half day reduction in length of stay for any intervention compared to control, but clinical and statistical heterogeneity was also identified in the network. Modelling intervention specific effects suggested beneficial effects of `procedural information’ and `sensory information’ in combination with `behavioural instruction’ or relaxation’. However, accounting for different effects of treatments did not explain heterogeneity. A proportion of the heterogeneity was explained by an association between the treatment effect and the mean LOS in the control group. Discussion: This analysis demonstrates an effective, rarely used approach to teasing apart the impact of intervention components on outcomes in evidence synthesis and the information gained from increasingly complex statistical models. The results obtained inform both theory and practice of psychological preparation for surgery.