Propensity Score vs Prognosis Summary Score: Two Similar-Looking Tools That Answer Different Questions
The Thai tailor-made Article is better
Two Similar-Looking Tools That Answer Completely Different Questions
In observational research, we often say we want to “make groups comparable.” But a deeper question comes first:
Comparable in what sense?
Are we trying to make two groups similar in terms of who would receive treatment? Or are we trying to make them similar in terms of what their outcome should look like at baseline?
This distinction leads us to two powerful but fundamentally different tools:
- Propensity Score (PS)
- Prognosis Summary Score (PSS)
At first glance, they look very similar. Both reduce many variables to a single number per person. Both can be used for matching or adjustment.
But underneath, they answer completely different clinical and methodological questions.
The Core Difference
- Propensity Score: “How likely is this person to receive treatment?”
- Prognosis Summary Score: “If this person were healthy, what should their outcome Y be?”
That’s it. Everything else follows from this difference.
Part 1: Propensity Score — Matching on Treatment Probability
What is a Propensity Score?
Propensity Score is the probability that a person receives a treatment (or exposure), based on their observed characteristics.
For each individual, we estimate:
“Given their age, sex, comorbidities, and clinical profile — how likely were they to be treated?”
How is it built?
You use all subjects (treated and untreated) and fit a model like:
- Treatment ~ age + sex + comorbidities + severity + etc.
This model learns how treatment decisions are made in real-world data.
Each person then gets a number between 0 and 1:
- 0.80 → very likely to be treated
- 0.20 → unlikely to be treated
How does matching work?
You match people with similar probabilities of receiving treatment.
Example:
- Patient A (treated) → PS = 0.72
- Patient B (untreated) → PS = 0.70 → match
- Patient C (untreated) → PS = 0.25 → not a match
What does this achieve?
After matching:
“These two people had the same chance of being treated, based on their characteristics.”
So any difference in outcome is less likely to be due to:
- disease severity
- comorbidities
- selection bias
This is especially important in situations like:
- Confounding by indication (sicker patients get treated)
- Confounding by contraindication (frail patients avoid treatment)
Key idea
Propensity Score balances treatment assignment, not outcome.
It answers:
“Are these two people equally likely to receive treatment?”
Part 2: Prognosis Summary Score — Matching on Expected Outcome
Now we shift perspective.
Instead of focusing on treatment, we focus on outcome.
What is Prognosis Summary Score?
PSS answers a very intuitive clinical question:
“If this person were healthy — what should their outcome be?”
Or in plain terms:
“Given their age, sex, BSA, HR — what would a normal value look like for them?”
How is it built?
Here’s the critical difference:
👉 You build the model using controls only (healthy individuals)
Example:
- Outcome (e.g., CMR value) ~ age + sex + BSA + HR + etc.
This model learns normal physiology.
Then what?
You apply this model to everyone:
- controls → expected ≈ observed
- cases → expected ≠ observed
Each person gets a predicted value:
“This is what their outcome should be if they were healthy”
That predicted value is the PSS
How does matching work?
You match people based on expected outcome, not observed outcome.
Example:
👉 A and B match because:
- Expected values are similar (110 vs 108)
- Not because observed values are similar
What does this achieve?
After matching:
“These two people should have had the same outcome if they were healthy.”
So any difference we observe is more likely due to disease effect.
Key idea
PSS balances baseline physiology, not treatment.
It answers:
“Should these two people have the same outcome under normal conditions?”
Why This Difference Matters
Let’s compare directly:
A Simple Clinical Analogy
Imagine studying exam scores.
Propensity Score approach:
Match students who had the same chance of getting tutoring
PSS approach:
Match students who should have had the same expected exam score based on ability
If your question is:
“Does tutoring improve scores?”
→ Propensity Score makes sense
If your question is:
“How much does a disease alter expected performance?”
→ PSS is more aligned
Why PSS is Powerful in Physiologic Studies
In fields like imaging, cardiology, or biomarker research, we often care about:
“How much does disease shift someone away from normal?”
Not just:
“Are cases different from controls?”
PSS allows you to say:
- Not just “higher than control.”
- But “higher than expected for this person.”
This is a much more precise and clinically meaningful statement.
Common Mistakes
1. Thinking PS predicts outcome
It doesn’t. It only models treatment assignment.
2. Using observed Y in PSS matching
Wrong. You must use predicted Y (expected value)
3. Building PSS with cases included
This contaminates the “normal model” with disease effects
4. Choosing a method without thinking about the question
The method should follow the question — not the other way around
Final Takeaway
Both methods simplify complex data into one number per person.
But:
- Propensity Score makes people comparable in terms of treatment probability
- PSS makes people comparable in terms of expected outcome
So before choosing a method, ask:
“Am I trying to control how treatment was assigned? or Am I trying to understand how far someone deviates from normal?”
Your answer determines everything.