| Neuropsychological testing and screening for Alzheimer's disease is time consuming and demanding for the patient. Furthermore, tests have limited use in detecting early AD, which is under-recognized. The objective of this research project is to develop and calibrate an item bank for the efficient testing and screening for AD. A calibrated item bank is a large collection of neuropsychological test questions for which the measurement properties (difficulty, discrimination) are known. The calibration of the item bank will be carried out using item response theory statistical analyses (IRT) and is aimed to place all banked items on a common difficulty scale using log-odds as a unit of measurement. A calibrated item bank can be used to test subjects via a computer adaptive test (CAT) procedure. By using CAT, Dr. Robert Lindeboom and his group expect to reduce the test burden by 70% without loss in accuracy. Also, an item bank enables the identification of AD sensitive test questions to optimize screening for AD. Furthermore, users of different tests that have items in common with the item bank can compare their findings. Finally, both test questions and subjects are reported on the same log-odds scale. Log-odds can be transformed to the probability of a correct response on a given test question. Lindeboom et al. expect that the conversion of raw test scores (e.g. '2 points' on the ADAS-COG) to log-odds units leads to a more straightforward interpretation of cognitive changes in patients than is now possible. In summary, with the development of an IRT calibrated item bank Lindeboom and his group want: - Reduce the test burden of eligible subjects with memory complaints. - Earlier recognition of subjects with probable AD than is now possible. - Improved communication of research findings. - Straightforward interpretation of test results because of the use of a formal epidemiologic measurement unit, the log-odds, which may be helpful in guiding therapeutic decisions. |
