Rasch Analysis of a Behavioral Checklist for the Assessment of Pain in Critically Ill Adults

Patients hospitalized in the intensive care unit (ICU) are often unable to report their pain, which is a problem since untreated pain is associated with negative health outcomes. The use of behavioral pain scales are recommended for the detection of the presence of pain in this vulnerable population. Previous validation studies have used classical techniques, and several psychometrics properties remain unknown. In this paper, data obtained from a behavioral checklist of dichotomized items was utilized to evaluate the instrument's dimensionality, its construct validity and its capacity to distinguish between levels of pain by using Rasch measurement. A sample of 239 ICU patients was used to collect the data. Results showed that, while unidimensionality was acceptable, concerns remained about the local independence and item fit indices. A third of the items showed misfit. Finally, while items had a great reliability (0.97), persons' measures had a rather low reliability (0.62) and only 1.28 strata of pain could be distinguished. The narrow range of pain levels in the sample could explain this poor performance and further study is needed, with a sample exhibiting a wider range of pain levels.