Methods to handle common data problems for longitudinal hidden Markov models are presented. A missing data mechanism that assumes state-dependent and variable dependent missingness is introduced. High dimensionality is controlled for with the use of an explicit dimension reduction algorithm.