Propensity Scores: A Practical Method for Assessing Treatment Effects in Pain and Symptom Management Research Melissa M. Garrido, PhD Journal of Pain and Symptom Management Volume 48, Issue 4, Pages 711-718 (October 2014) DOI: 10.1016/j.jpainsymman.2014.05.014 Copyright © 2014 Terms and Conditions
Fig. 1 Propensity scores should include hypothesized strong confounders and weak confounders, but they should not include instrumental variables (Guidelines 1–3). Journal of Pain and Symptom Management 2014 48, 711-718DOI: (10.1016/j.jpainsymman.2014.05.014) Copyright © 2014 Terms and Conditions
Fig. 2 Covariates included in a propensity score should not be affected by the treatment variable. Variables that occur after the treatment variable are part of the causal pathway; including these in the propensity score will bias the treatment effect one is trying to estimate (Guideline 4). Journal of Pain and Symptom Management 2014 48, 711-718DOI: (10.1016/j.jpainsymman.2014.05.014) Copyright © 2014 Terms and Conditions