Objective: To determine the feasibility, the cluster design effect and the variance and minimal clinical importance difference in the primary outcome in a pilot study of a structured approach to goal-setting.
Design: A cluster randomized controlled trial.
Setting: Inpatient rehabilitation facilities.
Subjects: People who were admitted to inpatient rehabilitation following stroke who had sufficient cognition to engage in structured goal-setting and complete the primary outcome measure.
Interventions: Structured goal elicitation using the Canadian Occupational Performance Measure.
Main measures: Quality of life at 12 weeks using the Schedule for Individualised Quality of Life (SEIQOL-DW), Functional Independence Measure, Short Form 36 and Patient Perception of Rehabilitation (measuring satisfaction with rehabilitation). Assessors were blinded to the intervention.
Results: Four rehabilitation services and 41 patients were randomized. We found high values of the intraclass correlation for the outcome measures (ranging from 0.03 to 0.40) and high variance of the SEIQOL-DW (SD 19.6) in relation to the minimally importance difference of 2.1, leading to impractically large sample size requirements for a cluster randomized design.
Conclusions: A cluster randomized design is not a practical means of avoiding contamination effects in studies of inpatient rehabilitation goal-setting. Other techniques for coping with contamination effects are necessary.