Hazard Ratios and Interpretation
You'll learn: How to interpret exponentiated Cox model coefficients as hazard ratios that quantify how covariates multiply the risk of an event.
What's a Hazard Ratio?
In the Cox Proportional Hazards Model, raw coefficients aren't intuitive. When you exponentiate them (raise e to their power), you get hazard ratios (HR) — much easier to understand!
A hazard ratio tells you the multiplicative change in the hazard (risk of the event at any moment) when a covariate increases by one unit, holding other variables constant.
Reading Hazard Ratios
- HR = 1: No effect. The covariate doesn't change the hazard.
- HR > 1: Increased hazard. For example, HR = 2.0 means the hazard doubles (200% of baseline).
- HR < 1: Decreased hazard (protective effect). HR = 0.5 means the hazard is halved (50% of baseline).
An Intuitive Example
Imagine modeling customer churn with age as a covariate. If the coefficient for age is 0.05, the hazard ratio is e^(0.05) ≈ 1.05.
Interpretation: Each additional year of age multiplies the hazard of churning by 1.05, meaning a 5% increase in churn risk per year.
If the coefficient were -0.22, the HR would be e^(-0.22) ≈ 0.80 — each extra year reduces churn hazard to 80% of what it was (a 20% decrease).
Continuous vs Categorical Variables
For continuous variables, the HR reflects a one-unit change. For binary variables (e.g., treatment vs control), the HR directly compares the two groups' hazards.
Key Takeaway: Hazard ratios express how covariates multiply the instantaneous risk of an event — values above 1 increase risk, below 1 decrease it, making Cox model results actionable and interpretable.