Shift work and sex are variants across Karolinska Sleep Questionnaire and the application of linking analysis as a step forward using Item Response Theory.

The frequency of sleep disturbances is considerably higher in the night and shift workers and in females than in day workers and males, respectively. However, a subjective sleep scale must be invariant across these groups, independently of the level of their members on the scale. This study is aimed to test the invariance of the Karolinska Sleep Questionnaire's (KSQ) items by shift work and sex. We used the data from a census that covered more than 90% (N = 1648) of the nurses from the main institute of the largest public hospital complex of Brazil. Firstly, we intend to find the KSQ's dimensionality using factorial analysis and Item Response Theory (IRT) performed by Graded Response Model. Differential Item Functioning (DIF) was the technique used to test the invariance of each KSQ's dimensions. In case of variance detection, we applied the linking analysis. Intending to test the KSQ's consistency with external variables, we assessed correlations between KSQ's dimensions with health variables, i.e., self-reported health status and musculoskeletal pain. We have found one scale and two subscales from one general and another bidimensional factor structure of the KSQ, respectively. In these dimensions, the KSQ's items fitted well to the IRT and we have identified DIF by shift work. However, we have found DIF by sex just in one item on the general factor. Linking analysis showed as a possible step forward in the variance issue placing on to the same scale the shift work groups in the items with DIF. All correlations between KSQ's dimensions with health variables were significant. Our findings allow us to argue that KQS's items were variant by shift work and sex in a nursing staff census from the largest public hospital complex of Brazil, but we can go on using linking analysis. This could be used as an evidence for the construct validity should go beyond the traditional dimensionality assessment. The dimensionalities of KSQ fit well for other population but individuals living in Scandinavian countries.