Relative Risk Calculator

What Is Relative Risk?

Relative risk (RR), also called risk ratio, compares the probability of an outcome between two groups:

$RR = \frac{P(\text{outcome} \mid \text{exposed})}{P(\text{outcome} \mid \text{unexposed})}$

An RR of 2.0 means the exposed group is twice as likely to experience the outcome. An RR of 0.5 means half as likely.

The 2×2 Table

Relative risk is calculated from:

$RR = \frac{a/(a+b)}{c/(c+d)}$

Example:

RR = (30/200) / (10/300) = 0.15 / 0.033 = 4.5

People with high blood pressure are 4.5 times more likely to have a heart attack.

Interpretation

Statistical Significance

If the 95% CI includes 1.0, the RR is not statistically significant.

CI: [1.8, 3.2] — Significant (doesn't include 1) CI: [0.8, 2.5] — Not significant (includes 1)

Applications

Clinical Trials

"Treatment reduced risk of recurrence: RR = 0.70, 95% CI [0.55, 0.90]"

A 30% relative reduction in risk.

Cohort Studies

Following groups over time:

"Smokers had RR = 15 for lung cancer compared to never-smokers"

Public Health

Vaccine effectiveness:

"Vaccination reduced infection risk: RR = 0.25 (75% relative reduction)"

Epidemiology

Disease risk factors:

"Obesity increased diabetes risk: RR = 3.2"

Confidence Interval for RR

The $\log(RR)$ is approximately normally distributed:

$\ln(RR) \pm z^* \times SE(\ln(RR))$

Where:

$SE(\ln(RR)) = \sqrt{\frac{1}{a} - \frac{1}{a+b} + \frac{1}{c} - \frac{1}{c+d}}$

Then exponentiate to get CI on the RR scale.

Example:

ln(4.5) = 1.504 SE = √[1/30 - 1/200 + 1/10 - 1/300] = 0.374 95% CI for ln(RR): 1.504 ± 1.96(0.374) = [0.77, 2.24] 95% CI for RR: [e^0.77, e^2.24] = [2.16, 9.35]

RR vs. Odds Ratio

RR: Ratio of probabilities — more intuitive

OR: Ratio of odds — useful in case-control studies

When outcome is rare (< 10%): OR ≈ RR When outcome is common: OR overstates the effect

For common outcomes, report RR—it's more meaningful.

When to Use Each

Use RR when:

• Prospective cohort or experimental study
• You can calculate actual incidence/risk
• Clinical interpretation is important

Use OR when:

• Case-control study (can't calculate RR)
• Logistic regression
• Matched designs

Relative Risk Reduction (RRR)

$RRR = 1 - RR = \frac{\text{Risk}_{\text{control}} - \text{Risk}_{\text{treated}}}{\text{Risk}_{\text{control}}}$

If RR = 0.70, RRR = 0.30 = 30% reduction.

"The drug reduced risk by 30%."

Absolute Risk Reduction (ARR)

$ARR = \text{Risk}_{\text{control}} - \text{Risk}_{\text{treated}}$

If control risk = 10% and treatment risk = 7%:

• RR = 0.70 (30% relative reduction)
• ARR = 3% (3 percentage points absolute reduction)

ARR is often more clinically meaningful.

Number Needed to Treat (NNT)

$NNT = \frac{1}{ARR}$

If ARR = 3%, NNT = 33.

"You need to treat 33 patients to prevent one event."

NNT depends on baseline risk. Same RR gives different NNT for different baseline risks.

Historical Context

Relative risk became standard in epidemiology through the work of mid-20th century researchers.

Richard Doll and Austin Bradford Hill (1950s) used relative risk to demonstrate the smoking-lung cancer link, establishing it as the primary effect measure in cohort studies.

Jerome Cornfield clarified the relationship between RR and OR, showing when each was appropriate.

The measure gained prominence as prospective studies became more common in medical research.

Adjusted Relative Risk

Crude RR may be confounded. Statistical models provide adjusted estimates:

Cox regression (hazard ratio approximates RR for rare outcomes) Log-binomial regression (directly estimates RR) Poisson regression with robust variance (common approximation)

"Crude RR = 2.5; age-adjusted RR = 1.9"

The adjusted RR controls for confounding.

Relative Risk in Subgroups

RR may vary across subgroups (effect modification):

Overall RR = 2.0

• In men: RR = 2.8
• In women: RR = 1.4

Reporting overall RR may mask important differences.

Assumptions and Limitations

Requires Prospective Data

RR needs incidence data—you must follow people over time or have true population risks. Case-control studies can't provide RR.

Depends on Comparison Group

RR = 4 comparing heavy smokers to never-smokers is different from RR = 2 comparing heavy to light smokers.

Ignores Baseline Risk

RR = 2.0 could mean:

• From 1% to 2% (small absolute increase)
• From 40% to 80% (huge absolute increase)

Always consider absolute risks.

Time Period Matters

"5-year RR = 2.0" differs from "10-year RR = 1.5"

Risks accumulate differently over time.

Related Calculations

• Odds Ratio Calculator — For case-control studies
• CI for Difference in Proportions — Absolute risk difference
• Confidence Interval for Proportion — Individual risks
• Standard Error of Proportion — Components of RR CI

Summary

Relative risk compares the probability of an outcome between exposed and unexposed groups. An RR of 1 indicates no difference; CIs excluding 1 indicate statistical significance. RR provides intuitive interpretation—"twice as likely"—and is the preferred measure for cohort studies and clinical trials. However, RR alone can be misleading without context about absolute risk. Always report both relative and absolute measures (RR, ARR, NNT) to give a complete picture of clinical significance.