The PROMIS of Better Outcome Assessment: Responsiveness, Floor and Ceiling Effects, and Internet Administration

Sample sizes required at 80% power. Power-sample size estimates for 3 questionnaires with 4, 6, and 8 items; the fewer-item sets contain subsets of the larger item sets. Data are from simulation studies setting the effect size to 0.2 and using a population with a mean physical function score 1 SD below the population mean (as with a population of moderately affected patients with RA), at the population mean, and 1 SD above the population mean. Excellent power is achievable with 8 (or more) items with the population below the mean. The improvement effect of increasing item numbers is illustrated. However, when the population is 1 SD above the population mean, power is poor and sample size requirements large. This is the effect of studying a population where many subjects are near the ceiling where few items conveying little information are available. The increased statistical power available by adding items at the floor and ceiling to existing PROMIS item banks is inferred.