Objectives: The primary aim of treatments for COPD is to improve health-related quality of life. However, little is known of the clustering effects related to health-related quality of life as an outcome measure. If clustering effects are observed, these have important implications for sample size estimates when cluster randomization is used in interventional studies. This study aimed to determine the intracluster correlation coefficient (ICC) of the quality of life, between hospitals for COPD patients.
Methodology: The Dyspnoea Impact and Symptoms Questionnaire was administered to 100 COPD inpatients from four public hospitals (25 from each) around metropolitan Melbourne, selected on the basis that they had not had any major programs implemented within the last 2 years that aimed to improve the management of COPD. Data were collected concerning demographic and socioeconomic variables and comorbidities.
Results: The highest ICC value for a health-related quality of life subscale was 0.02 (psychological score), while the highest for a symptom-based subscale was 0.04.
Conclusions: There is minimal clustering effect of quality of life in COPD patients between the hospitals studied. Despite this, when using a cluster randomised design the sample size needed to detect the same effect as a study using simple randomisation could be inflated by up to 183%. If cluster randomization is required, the average cluster size should be kept as small as possible to negate this effect.