The reliability of stroke outcome measures

Assessments of outcome in stroke should include measures of the broader aims of rehabilitation, such as handicap, mood and perceived health, and these measures should be reliable. This study compared the test-retest reliability of the Barthel, Nottingham Extended Activities of Daily Living, Geriatric Depression Score (GDS) and the Nottingham Health Profile (NHP) by postal survey of 21 stroke patients six months after their stroke. There was significant variation between test and retest scores in all the scales but the NHP and GDS were much less reliable than the disability measures, perhaps because they measure less objective or more intrinsically variable quantities. Unreliability is important because it adds to the observed variance of any study population, making it more difficult to detect real differences between groups. Analysis of test-retest reliability should be an essential part of the development of questionnaires. Currently recommended outcome measures require further development to improve reliability.