This JAMA Guide to Statistics and Methods reviews how propensity score methods can be used with observational data to mimic comparison populations in a randomized trial and account for differences that might lead to biased conclusions.
In a propensity score–matched cohort study published in the March 12, 2019, issue of JAMA, Zeng et al1 found that prescription tramadol was associated with significantly greater 1-year mortality compared with nonsteroidal anti-inflammatory alternatives in adults with osteoarthritis. At baseline, patients receiving tramadol were different than those who received other analgesics in terms of demographics, medical comorbidities, medications, and prior hospital resource utilization. Zeng et al1 used propensity score matching in an effort to account for differences between groups.2 This matched sample corresponds to a unique target population.
EXPLANATION OF THE CONCEPT
What Is a Target Population?
The target population is an intended group of patients characterized by inclusion and exclusion criteria and described by baseline characteristics to whom the average treatment effect applies. Two samples derived from a cohort with the same inclusion and exclusion criteria can have different characteristics that represent different target populations because of variation in sites and the way patients were enrolled in the study (Box 1).
Box 1 Target Population, Propensity Score Weighting, and Propensity Score Matching
The target population, to whom the average treatment effect generalizes, depends not only on inclusion and exclusion criteria, but also on how patients enter the sample. In observational analyses this is further influenced by the choice of propensity score method.
Propensity score weighting, using the form of inverse probability of treatment weighting, applies to all sampled patients who received either comparator (A or B), but may excessively weight findings from patients who would typically receive only 1 of these options and are not good candidates for the other. Propensity score weighting addresses the question, “What if everyone in the sampled population received one treatment vs if everyone received the other treatment?”
Propensity score matching in observational studies can emulate the populations in randomized clinical trials by having relatively narrow patient characteristics that typify patients who are most eligible for either treatment being compared, but usually exclude a portion of the sample. Propensity score matching typically addresses the question, “What if everyone in the sampled population who could be matched received one treatment or the other treatment?”
Why Is the Target Population Important?
Understanding a study’s target population is important for knowing how study results apply to certain types of patients. Observational studies performed with propensity score methods can change the target population by shifting the distribution of patient characteristics that contribute to analysis. Thus, propensity score analyses are used to reduce bias in the comparison between a population that received treatment and a control population and effectively mimic different randomized clinical trials (RCTs) ...