CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1080
Title: Examining service utilisation among the frail elderly using latent class analysis Authors:  Arthur White - Trinity College Dublin (Ireland) [presenting]
Jason Wyse - Trinity College Dublin (Ireland)
Abstract: Data collected by the Irish Longitudinal Study on Ageing (TILDA) are analysed. This is a prospective cohort study which is representative of the elderly Irish community dwelling population. Our analysis focuses on respondents aged 65 years and over who are deemed to be frail, with variables related to the use of services such as dentistry, informal care, and hospital stay considered. Some of these variables are binary, indicating presence or absence of use, while others are integer valued, and relate to frequency of use. While it is suspected that heterogeneity of service use and need exists among the elderly frail in Ireland, its extent, and the services for which it is most pronounced, are poorly understood. To examine this issue, a latent class analysis model is developed and fitted to the data. This consists of a mixture of binomial and Poisson distributions that correspond to the relevantly valued variables. Methods for model and variable choice are examined, and the broader implications of these choices and the resultant clusterings are discussed.