Purchase this article with an account.
Sheela Evangeline Kumaran, Nicole Ann Carnt, Sing Ling Stephanie Yeo, Konrad Pesudovs; Application of Rasch model to score the SF-36 instrument administered to a sample with acanthamoeba keratitis. Invest. Ophthalmol. Vis. Sci. 2021;62(8):3594.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The medical outcomes short form (SF-36) survey is widely used to measure generic health related quality of life (QoL). The survey covers eight key health concepts and is traditionally scored using summative scoring technique. The aim of this study was to apply Rasch analysis, a modern probabilistic measurement model to score the SF-36 survey that was administered to a clinical sample of contact lens wearers with acanthamoeba keratitis (AK).
The SF-36 survey was completed by 77 AK patients from Moorfields Eye Hospital, UK. Ordinal responses to the SF-36 items were coded such that higher value indicate better QoL. The Andrich rating scale Rasch model was applied using the Winsteps program (version 4.4.8). The psychometric properties (category functioning, measurement precision, fit statistics, and dimensionality) of the original SF-36 subscales and Rasch-derived subscales were explored and person measures (scores) corresponding to valid subscales were obtained.
The median age (range) of the participants was 33.6 years (19.9 –76.5) and 42 (54.5%) were female. Of the eight original subscales, three, namely the physical functioning (PF), role limitations due to physical health (RLPH) and role limitations due to emotional problems (RLEM) did not form valid scales due to low measurement precision (person separation index <1). Re-analysis starting with all 36 items in the model and testing for principal components yielded five valid scales consistent with the original design: role limitations (9 items), general health (5 items), pain (2 items), emotional well-being (5 items) and energy/fatigue (4 items). A newly organized “role limitations” scale encompassed items from the original SF-36 subscales of RLPH, RLEM and social functioning. The physical functioning items did not form a valid scale and were eliminated. The final Rasch-derived scales exhibited optimal category functioning, unidimensionality and excellent reliability (Wright’s sample independent person /test reliability > 0.85). These psychometrically valid, interval scales were able to generate meaningful person measures.
Five psychometrically valid SF-36 scales were derived using Rasch analysis for this sample with AK. These could be used to assess the burden of AK disease in contact lens wearers and potentially in other eye conditions.
This is a 2021 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only