Reliability of the Barthel Index when used with older people

OBJECTIVE: the Barthel Index (BI) has been recommended for the functional assessment of older people but the reliability of the measure for this patient group is uncertain. To investigate this issue we undertook a systematic review to identify relevant studies from which an overview is presented. METHOD: studies investigating the reliability of the BI were obtained by searching Medline, Cinahl and Embase to January 2003. Screening for potentially relevant papers and data extraction of the studies meeting the inclusion criteria were carried out independently by two researchers. RESULTS: the scope of the 12 studies identified included all the common clinical settings relevant to older people. No study investigated test-retest reliability. Inter-rater reliability was reported as 'fair' to 'moderate' agreement for individual BI items, and a high percentage agreement for the total BI score. However, these findings were difficult to interpret as few studies reported the prevalence of the disability categories for the study populations. There may be considerable inter-observer disagreement (95% CI of +/-4 points). There was evidence that the BI might be less reliable in patients with cognitive impairment and when scores obtained by patient interview are compared with patient testing. The role of assessor training and/or guidelines on the reliability of the BI has not been investigated. CONCLUSIONS: although the BI is highly recommended, there remain important uncertainties concerning its reliability when used with older people. Further studies are justified to investigate this issue.