Improving the use of propensity score methods in health research

Scientists frequently want to estimate the effects of treatments, interventions and exposures on patient outcomes (e.g., the effect of smoking on heart disease or the effects of Covid-19 vaccines on variants of Covid-19). However, treated or exposed subjects often differ systematically from untreated or unexposed subjects (e.g., smokers may have diets and exercise habits that differ from non-smokers). For this reason, advanced statistical methods must be used to account for these differences in characteristics, so that the true effect of the treatment or exposure can be estimated. One such statistical method is known as the propensity score. It allows for the comparison of subjects who resemble one another, apart from the treatment or exposure received. We will extend and improve methods based on the propensity score. This will allow the method to be applied in a wide variety of settings, including many that occur in health research.