Propensitetsskorning

Propensitetsskorning, or propensity scoring, is a statistical technique used in observational studies to reduce bias from confounding variables. When researchers cannot conduct a randomized controlled trial, they try to mimic randomization by creating comparable groups of subjects. Propensity scores help achieve this by estimating the probability that a particular individual will receive a certain treatment or intervention, given a set of observed covariates.

The process involves building a statistical model, typically logistic regression, where the outcome is the treatment assignment (e.g., receiving the drug vs. not receiving it) and the predictors are the observed confounding variables (e.g., age, sex, disease severity). The output of this model is a propensity score for each individual, representing their likelihood of receiving the treatment.

Once propensity scores are calculated, they can be used in several ways to adjust for confounding. Common methods include matching individuals with similar propensity scores across treatment groups, stratifying subjects into groups based on their scores, or using the scores as a covariate in regression analysis. The goal is to create a situation where the treatment and control groups are as similar as possible on observable characteristics, thus minimizing the impact of these confounders on the estimated treatment effect. Propensity scoring is widely used in fields like medicine, public health, and economics to draw more reliable causal inferences from non-randomized data.