Observational studies in the absence of selection bias reflect real life practices and have fewer constraints (ethical concerns, feasibility) than randomised clinical trials. Propensity score methods are increasingly used in observational studies to adjust for confounding and perform causal inferences (14 English Pubmed occurrences in the title or the abstract in 2000, 448 in 2010 and 3388 in 2018). This method allows interpretations of effects to be closer to causal inference like in a randomised clinical trial. There is a lack of articles describing propensity scores assumptions and interpretation rules. This leads to inadequacies in their use and in their reporting in scientific articles.We propose in this article a pragmatic synthesis, for the use of clinicians, presenting the advantages and limitations associated with the use of propensity scores for causal inferences. Our objective is to give clinicians keys to understanding and interpret propensity scores. We will develop an example based on simulated data to demonstrate the creation of the propensity score, four of its uses for covariate adjustment (matching, stratification, inverse probability weighing and adjustment) and its validation as well as reporting guidelines to ensure reproducibility of results in a scientific article.