Health Data Interpretation

The Numbers That Save or Scare You

A headline reads: 'New drug reduces stroke risk by 36%!' This sounds dramatic. But consider: if the baseline risk of stroke is 3%, a 36% relative reduction brings it to about 1.9%. The absolute risk reduction is 1.1 percentage points. Both numbers are correct. The first sounds like a breakthrough; the second sounds modest.

Pharmaceutical companies and health journalists almost always report relative risk reductions because they sound larger. Understanding the difference between relative and absolute risk is perhaps the single most important health data literacy skill. The metric that matters for your personal decision is the absolute risk — how much your actual probability of getting sick changes.

Number Needed to Treat (NNT)

The 'number needed to treat' (NNT) is the most patient-friendly way to express a drug's effectiveness. It answers: how many people need to take this drug for one person to benefit? Statins for heart attack prevention have an NNT of about 100 for low-risk patients — meaning 100 people must take statins daily for five years for one heart attack to be prevented. The other 99 get no benefit but may experience side effects.

Compare this to antibiotics for bacterial pneumonia, where the NNT is close to 1 — nearly every treated patient benefits. Or screening mammography, where the NNT to prevent one breast cancer death over 10 years is approximately 2,000 for women aged 50-59.

NNT does not mean a drug is useless — if you are the one person whose heart attack is prevented, the benefit is enormous. But it enables informed decision-making by making the actual probability of benefit transparent.

A Worked Example: Relative vs Absolute Risk

Return to the '36% stroke risk reduction' headline and put numbers to it.

Relative risk reduction = (3.0 - 1.9) / 3.0 = 36.7%. Absolute risk reduction = 3.0% - 1.9% = 1.1 percentage points. The Number Needed to Treat = 1 / 0.011 = about 91 — roughly 91 people must take the drug for five years to prevent one stroke. All three numbers describe the same study. The '36%' is accurate but emphasizes the largest-sounding framing; the NNT of 91 tells you the practical odds that the drug helps any individual patient. When you see only a relative-risk headline, ask for the absolute numbers before deciding what it means for you.