Authors’ reply to Yamana and colleaguesBMJ 2015; 351 doi: https://doi.org/10.1136/bmj.h6761 (Published 14 December 2015) Cite this as: BMJ 2015;351:h6761
- Shiro Tanaka, associate professor1,
- Koji Kawakami, professor1
- 1Department of Pharmacoepidemiology, Graduate School of Medicine and Public Health, Kyoto University, Yoshida-Konoe-cho, Sakyo-ku, Kyoto 606-8501, Japan
We agree with Yamana and colleagues that it is impossible to rule out the effects of unmeasured confounders in our study.1 2 However, we think that the logic behind their critiques is flawed.
Firstly, they say that the model for calculating the propensity score is inadequate because the discriminatory ability, also called the C statistic, was not measured and propensity score distributions were identical …