Determinants of catastrophic and total health expenditures in Turkey: A modelling approach based on robust regression and random forest

This study aims to identify the determinants of household-level health expenditures in Turkey and to examine the conditions under which these expenditures reach a financially "catastrophic" level. Utilizing the Household Budget Survey Microdata Set provided by the Turkish Statistical Institute, the analysis comprehensively investigates the influence of socio-demographic and economic factors on health-related spending. A holistic methodological framework was developed by employing a range of modeling approaches, including classical logistic regression, robust regression, robust logistic regression, and the random forest algorithm, to predict health expenditures. Robust regression was used to identify the factors influencing total health expenditures, whereas classical logistic regression, robust logistic regression, and the random forest algorithm were applied to estimate and model catastrophic health expenditures. Among these methods, the random forest algorithm demonstrated the highest predictive performance for catastrophic health spending. Results from classical and robust logistic regression analyses indicated that households comprising low-income individuals, the unemployed, those with low levels of education, and individuals living alone are at a significantly higher risk of incurring catastrophic health expenditures. Furthermore, individual and structural factors such as smoking and the presence of elderly household members were also found to exert a significant influence on health spending