Item ordering in stochastically curtailed health questionnaires with an observable outcome

Abstract

A recent article proposed using stochastic curtailment to shorten health questionnaires that predict an observable outcome. The current paper investigates whether the efficiency gains resulting from this approach can be enhanced by judiciously ordering the items within a questionnaire. Several new statistical procedures for ordering items are introduced and compared with an existing item ordering procedure, as well as with random orderings. In a post-hoc simulation using data from the Medicare Health Outcomes Survey, the orderings based on statistical criteria exhibited larger efficiency gains than the random item orderings. Comparisons between the different statistical methods depended on the simulation condition studied. Practical considerations are discussed.