Julie Barber1, Simon Thompson. 1. Research and Development Directorate, University College London Hospitals R&D Directorate, 1st Floor, Maple House, 149 Tottenham Court Road, London, W1P 9LL, UK.
Abstract
OBJECTIVE: Choosing an appropriate method for regression analyses of cost data is problematic because it must focus on population means while taking into account the typically skewed distribution of the data. In this paper we illustrate the use of generalised linear models for regression analysis of cost data. METHODS: We consider generalised linear models with either an identity link function (providing additive covariate effects) or log link function (providing multiplicative effects), and with gaussian (normal), overdispersed poisson, gamma, or inverse gaussian distributions. These are applied to estimate the treatment effects in two randomised trials adjusted for baseline covariates. Criteria for choosing an appropriate model are presented. RESULTS: In both examples considered, the gaussian model fits poorly and other distributions are to be preferred. When there are variables of prognostic importance in the model, using different distributions can materially affect the estimates obtained; it may also be possible to discriminate between additive and multiplicative covariate effects. CONCLUSIONS: Generalised linear models are attractive for the regression of cost data because they provide parametric methods of analysis where a variety of non-normal distributions can be specified and the way covariates act can be altered. Unlike the use of data transformation in ordinary least-squares regression, generalised linear models make inferences about the mean cost directly.
OBJECTIVE: Choosing an appropriate method for regression analyses of cost data is problematic because it must focus on population means while taking into account the typically skewed distribution of the data. In this paper we illustrate the use of generalised linear models for regression analysis of cost data. METHODS: We consider generalised linear models with either an identity link function (providing additive covariate effects) or log link function (providing multiplicative effects), and with gaussian (normal), overdispersed poisson, gamma, or inverse gaussian distributions. These are applied to estimate the treatment effects in two randomised trials adjusted for baseline covariates. Criteria for choosing an appropriate model are presented. RESULTS: In both examples considered, the gaussian model fits poorly and other distributions are to be preferred. When there are variables of prognostic importance in the model, using different distributions can materially affect the estimates obtained; it may also be possible to discriminate between additive and multiplicative covariate effects. CONCLUSIONS: Generalised linear models are attractive for the regression of cost data because they provide parametric methods of analysis where a variety of non-normal distributions can be specified and the way covariates act can be altered. Unlike the use of data transformation in ordinary least-squares regression, generalised linear models make inferences about the mean cost directly.
Authors: Eric A Latimer; Daniel Rabouin; Zhirong Cao; Angela Ly; Guido Powell; Tim Aubry; Jino Distasio; Stephen W Hwang; Julian M Somers; Vicky Stergiopoulos; Scott Veldhuizen; Erica E M Moodie; Alain Lesage; Paula N Goering Journal: CMAJ Open Date: 2017-07-18
Authors: Catherine Keating; Martin Neovius; Kajsa Sjöholm; Markku Peltonen; Kristina Narbro; Jonas K Eriksson; Lars Sjöström; Lena M S Carlsson Journal: Lancet Diabetes Endocrinol Date: 2015-09-17 Impact factor: 32.069
Authors: Clio Spanou; Sharon A Simpson; Kerry Hood; Adrian Edwards; David Cohen; Stephen Rollnick; Ben Carter; Jim McCambridge; Laurence Moore; Elizabeth Randell; Timothy Pickles; Christine Smith; Claire Lane; Fiona Wood; Hazel Thornton; Chris C Butler Journal: BMC Fam Pract Date: 2010-09-21 Impact factor: 2.497