Abstract
The different propensity score estimators reflect the average effect on the different populations. Particularly, it is pointed out that different causal inference methods based on propensity scores lead to entirely different conclusions when the treatment effect is not uniform across the study population. However, many clinical studies did not care about the difference in the estimands. To illustrate the difference in the estimated values depending on the propensity score methods in practice, were-analyzed a case study assessing the effects of surgical treatment among tongue cancer patients, which the treatment effect varied depending on the patients’ characteristics. Then we conducted a computer simulation to verify the results of the case study.
Acknowledgments
We are grateful to Dr. Otsuru of Nagasaki University for sharing the case study data. Furthermore, we would like to express our deep gratitude to Professor H. Iso of Osaka University for carefully reading the article and giving many helpful suggestions.
Disclosure statement
No potential competing interest was reported by the authors.