Advised missing data method

Longitudinal analysis > Missings in predictor/covariate > MNAR > <50% missing data > >75% missing item scores

The missing data method should be applied to the total scores of the questionnaire directly. In this situation where the missing data are MNAR and no more than 50% of the subjects have missing data, multiple imputation can be performed. Multiple imputation was developed to work on MAR data, but was shown to perform well in MNAR data when the percentage of subjects with missing values is not too large (Eekhout et al. 2014).