Premium
P1‐213: Stated preference methods in questionable dementia evaluation: an emerging methodological paradigm in health economics
Author(s) -
Prolo Paolo,
Posa Franco,
Luisoli Sergio
Publication year - 2009
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2009.04.220
Subject(s) - dementia , cognition , psychology , actuarial science , medicine , clinical psychology , cognitive psychology , psychiatry , business , disease , pathology
similar to those used in the AD Neuroimaging Initiative. We regressed DS scores (range 0-15) against CDR-SB (range 0-18) using simple linear, piecewise linear (spline), and polynomial (up to fifth order) models to investigate the strength and shape of the relationship (linear vs. non-linear) between the two measures. Model fit was evaluated using R and mean squared error (MSE). Results: Pearson correlation coefficients suggested a moderately high correlation between the two measures (r1⁄40.82430, p<0.0001). Simple linear regression performance (R1⁄466.9%, MSE1⁄42.84) was similar to more complex models, including a cubic polynomial model (R1⁄468.8%, MSE1⁄42.69) and a spline regression with two breakpoints (R1⁄469.3%, MSE1⁄42.66), suggesting that a constant relationship slope for the DS across the range of CDR-SB scores. The simple linear regression provided an estimate of DS 1⁄4 0.93 þ 0.68(CDR-SB), indicating a one point change on the CDR-SB corresponded to a change of 0.68 DS points. Conclusions: In subjects with MCI or mild-moderate AD, the DS and CDR-SB, while measuring unique constructs, are well correlated, providing an opportunity to compare results across different populations where one measure but not the other was used. More complex regression models only modestly improved fit, indicating the DS can be well described as a linear function of CDR-SB.