Published in Journal of Clinical Epidemiology, 2022
In clinical trials, patient-reported transition ratings ideally capture the change between a patient’s baseline health state and a present state. However, evidence suggests that these transition ratings may be more reflective of the present state than being balanced between the two. This is known as the Present State Bias (PSB) and it potentially undermines the validity or these transition ratings.
There have been several proposed criteria for determining the extent of PSB, and we wanted to assess how accurately these assess PSB. We also wanted to assess how well a Confirmatory Factor Analysis (CFA) approach performs in comparison. This CFA criterion was to be based on the ratio of the factor loadings of the transition ratings plus one.
For this study we simulated several samples with baseline and follow-up item responses to a
hypothetical questionnaire, and generated transition ratings. In addition, four real datasets were analyzed. We found that the samples varied according to their degree of PSB.
The performance of the proposed criteria and of our CFA-based approach, were evaluated based on how much PSB variance they explained. We found that the criteria proposed in the literature were able to explain between 36% and 74% of the variance in PSB. The new CFA-based criterion, however, explained between 81% and 98% of the PSB variance in the samples.