OBJECTIVES: This study attempts to investigate the impact of national mass screening program on health service utilizations and medical expenses. METHODS: Main data is consisted of the subjects (12,449,964) eligible for 2002 screening program provided by the National Health Insurance who were born in even year. To analyze the effect of screening in the case of hypertension, we excluded people who had any medical records of hypertension before 2002. After sorting and random sampling 5%(564,443), we extracted the subjects' medical payments for hypertension complications in 2007. The independent variable of interest is whether screened or not, and the dependent variables are the incidence and costs of hypertension complications. We compared the results obtained using instrumental variable methods with those from conventional logistic and linear regression models. RESULTS: Conventional logistic and multiple regression models suggest that the possibility of utilizing medical services is significantly higher in no-screen group and the average medical cost of hypertension complications reduced by 5% in screen group at 10% significant level. However, the results estimated using instrumental variable methods show different results. The screen group has significantly lower possibility to use medical service (p<0.05), but the lower average costs in screen group is not significant any more. CONCLUSIONS: The different results demonstrate that conventional regression approaches may have limitations in making causal inference using non-experimental data. This study shows that whether one gets medical screen or not affects the possibility of hypertension complications occurring, however it doesn't significantly relate to average costs.