Stochastic curtailment of health questionnaires: a method to reduce respondent burden

Abstract

Stochastic curtailment is a sequential method to terminate a study when continuing to the end would be unlikely to change the outcome. This method has been researched most commonly in the context of clinical trials. The current paper explores its use in a different setting: the administration of a health questionnaire to patients via computer. A classification procedure augmenting logistic regression with stochastic curtailment is introduced to avoid burdening the patients with unnecessary questions. In a real-data simulation using responses from the Medicare Health Outcomes Survey, the new procedure substantially reduced the average number of questions administered with a minimal loss of classification accuracy.