The Proportional Hazards Assumption
What you'll learn: You'll understand what it means for hazard ratios to remain constant over time—the core assumption behind Cox models.
The Core Idea
The proportional hazards assumption states that the ratio of hazards between any two groups (or for different covariate values) remains constant across all time points. In other words, if Group A has twice the hazard of experiencing an event compared to Group B at the beginning of the study, that 2:1 ratio should hold throughout the entire follow-up period.
Think of it like a race where two runners maintain the same relative speed advantage throughout. If Runner A is always 20% faster than Runner B—whether at mile 1 or mile 20—that's "proportional." The absolute speeds might change (they both slow down), but their ratio stays the same.
Why This Matters
The Cox Proportional Hazards Model assumes this proportionality. If the assumption holds, you can summarize the effect of a covariate (like treatment vs. control) with a single hazard ratio that applies to all time points. This makes interpretation clean and powerful.
However, if hazards cross over time—say Treatment A is better early on but Treatment B becomes superior later—the proportional hazards assumption is violated. A single hazard ratio wouldn't tell the full story.
Example in Plain Terms
Imagine comparing survival between smokers and non-smokers. If the proportional hazards assumption holds, the hazard of death for smokers might always be 1.5 times that of non-smokers, whether we're looking at 5 years or 20 years into the study. The absolute risk changes with age, but the ratio stays 1.5.
Key Takeaway: The proportional hazards assumption means the hazard ratio between groups remains constant over time—a critical requirement for interpreting Cox model results accurately.