An analysis of the costs of treating schizophrenia in Spain: a hierarchical Bayesian approach

Abstract

Background: Health care decisions should incorporate cost of illness and treatment data, particularly for disorders such as schizophrenia with a high morbidity rate and a disproportionately low allocation of resources. Previous cost of illness analyses may have disregarded geographical aspects relevant for resource consumption and unit cost calculation.

Aims: To compare the utilisation of resources and the care costs of schizophrenic patients in four mental-health districts in Spain (in Madrid, Catalonia, Navarra and Andalusia), and to analyse factors that determine the costs and the differences between areas.

Methods: A treated prevalence bottom-up three year follow-up design was used for obtaining data concerning socio-demography, clinical evolution and the utilisation of services. 1997 reference prices were updated for years 1998-2000 in euros. We propose two different scenarios, varying in the prices applied. In the first (Scenario 0) the reference prices are those obtained for a single geographic area, and so the cost variations are only due to differences in the use of resources. In the second situation (Scenario 1), we analyse the variations in resource utilisation at different levels, using the prices applicable to each healthcare area. Bayesian hierarchical models are used to discuss the factors that determine such costs and the differences between geographic areas.

Results: In scenario 0, the estimated mean cost was 4918.948 euros for the first year. In scenario 1 the highest cost was in Gava (Catalonia) and the lowest in Loja (Andalusia). Mean costs were respectively 4547.24 and 2473.98 euros. With respect to the evolution of costs over time, we observed an increase during the second year and a reduction during the third year. Geographical differences appeared in follow-up costs. The variables related to lower treatment costs were: residence in the family household, higher patient age and being in work. On the contrary, the number of relapses is directly related to higher treatment costs. No differences were observed between health areas concerning resource utilisation.

Discussion: Calculating the costs of a given disease involves two principal factors: the resource utilisation and the prices. In most studies, emphasis is placed on the analysis of resource utilisation. Other evaluations, however, have recognized the implications of incorporating different prices into the final results. In this study we show both scenarios. The factors that determine the cost of schizophrenia for the Spanish case are similar to the factors encountered in studies carried out in other countries.

Implications for health policies: Treatment costs may be reduced by the prevention of psychotic symptoms and relapse.

Implications for future research: The use of the same price data in multicentre studies may not be realistic. More effort should be made to obtain price data from all the centres or countries participating in a study. In the present study, only direct healthcare and social costs have been included. Future research should consider informal and indirect costs.