Missing data in a multi-item instrument were best handled by multiple imputation at the item score level

Publication
Journal of clinical epidemiology