Using the Rasch Model to Compare Medication Adherence Questionnaires

Abstract

Medication adherence questionnaires often provide heavily skewed results with limited variance, suggesting that most participants are highly adherent, contrasting with findings from objective adherence measures. We argue that one of the main limitations of the existing questionnaires is the limited range covered by the behaviours assessed. We have resolved this issue by developing a new questionnaire (the ProMAS) using IRT (the Rasch model). With use of the concept of ‘item difficulty’, items could be selected that better matched the behaviours performed. In the current study we compared the range of the non-adherence behaviours assessed in the ProMAS and the existing MARS. Elderly receiving medication for chronic conditions (N=370) completed both questionnaires. Results from the Rasch analysis indicated that the ProMAS covered a wider range of adherence behaviours than the widely used MARS. While the MARS only provided one item to distinguish between the 50% most adherent patients, the ProMAS provided six items. This resulted in adherence scores more in line with those obtained with objective adherence measures, showing more variance and less skewness. We conclude that the ProMAS is more capable of discriminating between people with different adherence rates than the widely used MARS, in particular in the high adherence range, illustrating the advantages of using IRT.