Heterogeneity in econometric analyses of health and health care -PhD Thesis-
This thesis deals with the issue of heterogeneity in econometric analyses of health and health care utilisation. Chapters 2 to 4 contain empirical analyses of health care utilisation using three different data sources with repeated observations for each individual. In these three chapters, a latent class panel data framework is adopted to model individual unobserved heterogeneity in a flexible way. Accounting for the panel structure of the data leads to a substantial improvement in fit, and permits the identification of latent classes of users of health care. In Chapter 2, using English data taken from the British Household Panel Survey for the period 1991-2001, a positive impact of income on access to and utilisation of primary care is found, especially amongst those with lower levels of utilisation. Chapter 3 proposes a new model that combines the features of the two most common approaches to econometric modelling of count measures of health care utilisation: the hurdle model and the finite mixture of Negative Binomial. The proposed finite mixture hurdle model is shown to fit the data substantially better than the existing models for an application to data on outpatient visits from the RAND Health Insurance Experiment. The results indicate a higher price effect for low users of health care, resulting mainly from the difference of the price effects on the probability to visit a doctor, rather than from those on the conditional number of visits. In Chapter 4, the finite mixture hurdle developed in Chapter 3 is applied to data on GP and specialist visits, from the European Community Household Panel. The finite mixture Negative Binomial and hurdle specifications outperform the standard hurdle model. Additionally, the finite mixture hurdle model reveals differences in the effect of income on the probability of use and the conditional number of visits. For specialist visits, low users are more income elastic than high users and the probability of using health care is more income elastic than the conditional number of visits.
Index to Theses search; not held in Res Lib - bibliographic reference only