Return to search

An explanation of the determinants of prescribing expenditure in the General Medical Services scheme in Ireland, with an application to budget setting

The Indicative Drug Target Scheme was introduced to the General Medical Services (GMS) scheme in 1993 with a view to encourage more rational and efficient prescribing. Each GP practice's drug budget is determined chiefly by the number of people on the GP's GMS panel adjusted for national age-related average prescribing expenditure. This thesis examines the factors associated with variations in GMS prescribing expenditure, proposes a number of alternative ways of constructing GMS drug budgets and examines the potential budgetary consequences. A unique dataset of individual-level and GP-level factors is constructed, including the first research application of an administrative database of demographic, socio-economic and access-related variables, the generation of chronic illness indicators, and the application of new measures of GP prescribing style. Multiple imputation of missing values and imputation of income are two additional innovations. Drawing on recent advances in risk-adjustment and the microeconometrics of health care utilisation, various specifications of an expenditure function are examined, given the skewed distribution of prescribing expenditure. These include logarithmic transformations, generalised linear models and finite mixture models. Quantile regression and outlier identification techniques are used for exploratory data analysis and to assist model specification. The principal determinants of prescribing expenditure are chronic illness, disability and age. Access to services and GP characteristics also have important effects. A number of competing models of budget setting are tested for predictive performance and distributive consequences. Most alternatives are an improvement on the current model. Compared to the current approach, the preferred model has greater explained variance, lower prediction error, is more pro-poor and has lower prediction error for vulnerable groups such as the disabled and the chronically ill.

Identiferoai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:726389
Date January 2003
CreatorsMcElroy, Brendan
PublisherUniversity of Edinburgh
Source SetsEthos UK
Detected LanguageEnglish
TypeElectronic Thesis or Dissertation
Sourcehttp://hdl.handle.net/1842/24939

Page generated in 0.0126 seconds