Multilevel modeling of regional variation in equity in health care

Background: The use of a concentration index is recommended to estimate socioeconomic equity in health and health services. Methods for the analysis of concentration indices have been developed in several studies. However, these methods do not take into consideration clustering within areas, which is necessary in a comprehensive study of regional variations in equity. Objectives: The study aims to develop a statistical method to assess variations in socioeconomic inequities in the use of health services in relation to need in different regions. Methods: Concentration index methods were developed further and the advantages of multilevel modeling were exploited. As an empirical example we analyzed revascularizations in 2001-2003 among the Finnish population. Results: The average inequity indices for the income distribution of revascularizations in Finland obtained with multilevel and standard regression modeling were comparable, but confidence intervals were smaller with multilevel modeling. Inequity indices for different areas estimated using multilevel modeling were more conservative and had smaller confidence intervals than indices estimated using the standard approach. Conclusions: The proposed approach is an efficient way of estimating regional variations in the socioeconomic equity of health care use. It enables the inclusion of need in the model and takes into account the varying need for services in different population groups and areas. In addition, the advantages of using multilevel modeling to estimate indices include the possibility to take into account dependence between observations within regions and to overcome the problems associated with random error in small regions.