Bootstrap corrected Propensity Score: Application for Anticoagulant Therapy in Haemodialysis Patients.

The inverse propensity score weighting (IPSW) has been often used to es- timate causal effects of treatments for observational data. However, IPSW requires strong assumptions, in which their misspecifications may severely bias estimated treatment effect. We present a bootstrap based bias-correction to adjust the propen- sity score weights in case of misspecifications of one of the main assumption. We showed, using simulation, the approach performs well in correcting biases due to model misspecifications in various contexts.The method was also illustrated using a real data based on end-stage renal disease.