An intrinsically valid approach to integrate several health outcomes into a comprehensive score
View Presentation
Tingting Song, The Rockefeller University
*Knut M. Wittkowski, The Rockefeller University
Keywords: nonparametric, multidimensional, rank, score, propensity
Assessing "health" and its determinants often requires integrating several ordinal and binary outcomes. In the absence of a "gold standard", traditional methods to justify a linearizing transformation and relative weight for each outcome fail, although "more" can be assumed to be "better". Recently, a nonparametric approach has been generalized to multivariate data. Uscores for multivariate data (muscores) are the first "intrinsically valid" approach, because they utilize only information that is invariant to (monotonous) scale transformations and (positive) weights. It is demonstrated how the "muStat" package (cran.rproject.org) and server (muStat.rockefeller.edu) can be applied to objectively integrate censored (timetoevent), longitudinal, and hierarchically structured (e.g., as risk, health, qualityoflife, cost) data, thereby providing a sound basis for policy decisions.
