Inter-rater reliability of postural observation after stroke

Abstract

Objective: To explore the inter-observer reliability of bedside observations of stroke patients’ posture using two versions of a pictorial tool. Design: Three projects were conducted. The initial version of the tool was used in project 1. The modified version was used in projects 2 and 3. In each project a pair of observers (comprising the main observer and one of five co-observers with varying degrees of experience in observing posture) used the tool to make simultaneous observations of 19 aspects of the posture of a sample of stroke patients. Each patient was observed in one or more of four positions (seated, supine and lying on the affected and unaffected side). The degree of inter-observer agreement was sought by calculating kappa values and percentage agreement. Setting: Medical wards, care of the elderly wards and a stroke unit. Subjects: A convenience sample of 57 stroke patients. Results: Four hundred and forty paired sets of observations were made (200 in project 1, 140 in project 2 and 100 in project 3). The main observer was in every pair. The co-observers made between 50 and 135 sets of observations each. When the results from all three projects were amassed, acceptable percentage agreement (i.e. 0%) was obtained for 67% ( n = 78) and 73% ( n = 55) of the results collected on aspects of the posture of the affected upper and lower limbs respectively. In contrast, acceptable percentage agreement for observations relating to the head, neck and trunk was obtained for only 34% ( n = 50) of the results collected. Uneven distributions in the data made kappa values difficult to interpret. Inter-observer agreement was not noticeably higher for pairs in which both observers had prior experience of observing posture after stroke than for pairs in which one observer was relatively inexperienced. Conclusions: The tool has potential as a quick and simple means of collecting information at the bedside about stroke patients’ posture. Refinements, additional training in using the tool for observers and further testing are suggested before its wider use is advocated.