How researchers mimic the 'gold standard' of medical trials using math
In short: In medicine, a randomized controlled trial is the gold standard for proving a treatment works. But what happens when you can't randomize—like when it would be unethical to assign someone to smoke or undergo a risky surgery? Scientists use propensity score matching to create a mathematical mirror of reality.
The fundamental challenge in observational medicine is selection bias. When researchers compare two groups—such as patients who received a new minimally invasive brain surgery versus those who underwent conventional resection—the groups are rarely equal at the start. Differences in age, health status, or previous conditions (confounders) can skew the results, making it impossible to tell if the outcome was caused by the treatment or by these pre-existing differences.
Propensity score matching (PSM) solves this by calculating the 'propensity' of a subject to receive a specific treatment based on their observed characteristics. Introduced by Paul R. Rosenbaum and Donald Rubin in 1983, the propensity score is essentially the conditional probability that a unit will be assigned to the treatment group given its covariates. By pairing subjects from both groups who have nearly identical scores, researchers can emulate the balance of a randomized controlled trial (RCT) using non-random data.
This technique allows for more precise causal inference rather than mere association. For example, in a study of gastroptosis, researchers used 1:1 propensity score matching to balance confounding factors between 188 patients with the condition and 376 controls. This balancing act ensures that when they identified predictors like body mass index or gastric motility, those findings were more likely tied to the condition itself rather than underlying differences in the study populations.
While PSM is a powerful tool for handling small sample sizes and complex observational data, it requires careful execution. Researchers must define the correct causal variables, estimate scores using models like logistic regression, and perform a 'balance check' to ensure the treatment and control groups are truly comparable before evaluating the final effect size.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Summary by Intellect · source: Propensity score matching for comparative studies: a tutorial with R and Rex
- Also: Propensity score matching
- Also: Clinical characteristics and early recognition of gastroptosis: a propensity score-matching analysis.
To cite this card, link to https://intellectplus.co.uk/app/health-propensity-score-matching.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
Related
More in Epidemiology · More in Health · Latest · A–Z · All topics
