Evaluation of Nursing Homes Using an Extended Latent Markov Model

We illustrate the use of an extended version of the latent Markov (LM) model to analyse data collected by the repeated administration of questionnaires aimed at evaluating nursing homes performance. In particular, we consider a longi- tudinal study on the health status of elderly people hosted in Italian nursing homes, named ULISSE. The questionnaire used to collect the data is made of a large num- ber of polytomous items. We consider only a subset of these items, selected through a suitable item selection algorithm. This subset is chosen so as to provide an amount of information close to that of the full set. Moreover, we propose an approach based on an LM model where both the initial and transition probabilities of the latent pro- cess depend on time-constant and time-varying covariates. The proposed model also allows us to account for missing responses on the basis of the assumption of latent ignorability. Aim of this approach is to estimate the effect of each nursing home on the probability of transition between latent states, which represent different level of the health status of the patients.