The Complete PSM Application Chain in Preventive Medicine
Engineering PSM for real-world preventive medicine requires solving three practical problems: data sourcing, matching quality assurance, and making results acceptable to payers.
Step 1: Data Integration and Preparation
PSM requires three data types: ①Baseline characteristics (age, BMI, blood pressure, lipids, smoking history, baseline risk score); ②Treatment assignment records (who received the intervention and when); ③Outcome metrics (health outcome changes during follow-up). All baseline variables must be measured before intervention begins.
Step 2: Calculate Propensity Scores and Execute Matching
Use logistic regression with all baseline covariates to predict each patient's probability of receiving the intervention — the propensity score. Apply 1:1 nearest-neighbor matching with caliper (typically 0.2 × SD of propensity score). After matching, verify all covariate SMDs are below 0.1.
| Metric | Pre-match | Post-match | Standard |
|---|---|---|---|
| Age SMD | 0.31 | 0.04 | <0.1 |
| BMI SMD | 0.24 | 0.06 | <0.1 |
| Baseline Risk SMD | 0.35 | 0.03 | <0.1 |
Step 3: Estimate ATT and Confidence Intervals
Calculate the outcome difference between matched treatment and control groups to obtain ATT (Average Treatment Effect on the Treated). Use Bootstrap resampling (≥1000 iterations recommended) for 95% confidence intervals.
PSM is not a single line of code — it is a data pipeline requiring rigorous quality control. Only PSM reports that pass a complete quality checklist constitute evidence acceptable for settlement negotiations.