ABSTRACT
Objective: This study aimed to identify the major variables in the estimation of health expenditure in OECD member countries with the decision tree method and to categorize the member countries by health expenditure
Methods: The study population comprised the 2014 data of the 35 OECD countries. In the study, health expenditure as a share of gross domestic product was the dependent variable while gross domestic product per capita, percentage of total population covered by public and private insurance, out-of-pocket health expenditure as percentage of total expenditure on health, age dependency ratio, life expectancy at birth, number of hospitals per million population, number of physicians per 1000 population/head counts, pharmaceutical sales and perceived health status were designated as independent variables. The decision tree model was constructed with the CART algorithm using the Orange data mining software package
Results: In the study, GDP per capita, life expectancy at birth, age dependency ratio, number of hospitals and percentage of the population with a bad perceived health status were identified as the major variables in the estimation of health expenditure. OECD countries were categorized in 6 groups according to the decision tree model. According to the CART algorithm used in the model, the classification accuracy rate and the precision of estimation were computed as 80.56% and 81.25%, respectively
Conclusion: The study results revealed that the most important determinant for estimating the share of GDP allocated to health expenditure was GDP per capita. Future studies should be conducted with the inclusion of different variables in the model