Comparison of the Selected Health Indicators of OECD Member Countries with Cluster and Discriminant Analysis

Objective: The aim of this study was to determine the usefulness of clustering and discriminant analysis to compare selected health indicators of OECD countries and common features among them. Material and Methods: The most significant variables among health indicators in 2004-health expenditures per capita, life expectation at birth and death number per 1000 births-were used in the analysis. Three methods were used for comparison during the clustering procedure, the hierarchical method and two non-hierarchical methods, the K-means and Medoid clustering method. The number of clusters was decided to be 3 according to the hierarchical clustering method. Results: According to the K-means analysis, all four health indicators were effective among the selected indicators of 30 countries (p<0.05). The Medoid clustering method revealed that the shadow statistics was 0.51, which meant that "there was acceptable clustering structure among the units". The separating analysis used for testing the accuracy of the clustering analysis revealed a high mean value, low Wilk's Lambda and high cononic correlation values. The study showed that the first 20 countries to join the OECD were developed with high incomes, transferring adequate amount of money to health expenses and Turkey was among the second ten countries to join the OECD with health indicators similar to the countries with above average income. Conclusion: In conclusion, according to hierarchical clustering, Turkey clusters with Poland, Slovak Republic, Czech Republic, Hungary, Mexico, Korean Republic, according to non-hierarchical (K-means) clustering, with Portugal, Poland, Slovak Republic, Hungary, Czech Republic, Mexico, and Korean Republic and according to Medoid clustering, only with Mexico.