|
Abstract:
|
Computerized adaptive testing (CAT) in clinical trials is an assessment method where an algorithm iteratively selects and administers potentially different items to each participant at each assessment. Items can measure a health concept such as physical functioning and are selected from a larger bank of items calibrated using an Item Response Theory (IRT) model to obtain parameters of location and discrimination for each item. CAT potentially provides individualized testing by selecting more appropriate items for a specific trial participant at a particular testing occasion. We will outline a few topics related to use of CAT to assess endpoints in clinical trials with regulatory intent. These include construction of item banks less prone to floor/ceiling effects, content coverage concerns, justification of the assumptions of the IRT model and CAT algorithm, interpretation of model estimates and scores for health-related concepts, the chance that CAT may not provide advantages over static assessments in terms of precision, number of items administered, and/or floor/ceiling effects.
|
