A Fresh Look at Out-of-Pocket Health Expenditures after More than a Decade Health Reform Experience in Turkey: A Data Mining Application
Strategies to combat with poverty are at the top of the agenda of Turkish health reform process. Health reform completed more than a decade and improvements in insurance coverage, incorporating green card scheme into the system are major parts of these reform process. There is a need to fresh look at the long term effects of this reform process on out-of-pocket (OOP) health expenditures and predictors of these expenditures. This study aims to fill this void by examining the trend of OOP health expenditures in Turkey from 2003 to 2015 and predictors of OOP health expenditures. Data came from Turkish Statistical Institute (TURKSTAT). Random Forest (RF) and Neural Network (NN) methods were compared to find predictors of OOP health expenditures for the year 2015 by generating a decision tree. RF outperforms NN and showed classification accuracy, sensitivity and specificity, value of 0.5352, 0.2925, 0.7855 respectively. The Area Under the ROC Curve (AUC) was 0.5539. Study results revealed that OOP health expenditures increased from 2003 to 2015. Moreover, education, marital status, being 65 years of age or older, income group and household size are important variables to predict OOP health expenditures respectively. This paper provides a current look at the effect of poverty alleviation strategies on OOP health expenses and emphasize that OOP health expenses have an increasing trend despite poverty reduction policies. New policies are needed to control increasing trend of OOP health expenditures in Turkey, it is hoped that this study will inspire health policy makers to fight against continuously increasing trend of non-zero OOP health expenditures in Turkey.
Keywords: Out-of-pocket health expenditures, poverty alleviation, Turkey, random forest, neural network
JEL Classifications: I10, I32, 053, C53, C53