Background: Equity in health is the focus of attention in the world health circles in recent decades. The financing of household health expenditure is, therefore, a concern in any region.

Many methods are available to analyze incomplete longitudinal ordinal responses. In this paper a general transition model is proposed for longitudinal ordinal responses with random dropout. Maximum likelihood estimates are obtained for the transition probabilities when there are repeated observations. The likelihood function of the general model is partitioned to make possible the use of existing software to estimate model parameters. Some reduced forms of the model are also considered where for estimation of parameters in these models one has to use numerical optimization methods. The approach is applied to the well-known Fluvoxamine data. For these data, two important results, which have not been previously reported, are obtained: (1) some transition probabilities are estimated to be zero and (2) the model for current response, which conditions on previous response, removes the effects of some covariates that had previously been significant.

In analyzing most correlated outcomes, the popular multivariate Gaussian distribution is very restrictive and therefore dependence modeling using copulas is nowadays very common to take into account the association among mixed outcomes. In this paper, we use Gaussian copula to construct a joint distribution for three mixed discrete and continuous responses. Our approach entails specifying marginal regression models for the outcomes, and combining them via a copula to form a joint model. Closed form for likelihood function is obtained by considering sampling weights. We also obtain the likelihood function for mixed responses where one of the responses, time to event outcome, may have censored values. Some simulation studies are performed to illustrate the performance of the model. Finally, the model is applied on data involving trivariate mixed outcomes on hospitalization of individuals, based on the survey of household's utilization of health services.

No abstract

There are many methods for analyzing longitudinal ordinal response data with random dropout. These include maximum likelihood (ML), weighted estimating equations (WEEs), and multiple imputations (MI). In this article, using a Markov model where the effect of previous response on the current response is investigated as an ordinal variable, the likelihood is partitioned to simplify the use of existing software. Simulated data, generated to present a three-period longitudinal study with random dropout, are used to compare performance of ML, WEE, and MI methods in terms of standardized bias and coverage probabilities. These estimation methods are applied to a real medical data set.