Mathwords logoMathwords

Odds Ratio — Definition, Formula & Example

An odds ratio is a single number that tells you how much higher or lower the odds of an outcome are in one group than in another. An odds ratio of 1 means the two groups have the same odds.

For a binary outcome and two groups, the odds of the outcome in a group are p1−p\frac{p}{1-p}, where pp is the probability of the outcome in that group. The odds ratio is the odds in group 1 divided by the odds in group 2: OR=p1/(1−p1)p2/(1−p2)\text{OR} = \frac{p_1/(1-p_1)}{p_2/(1-p_2)}. Its value is always positive. OR>1\text{OR} > 1 means the outcome is more likely in group 1, OR<1\text{OR} < 1 means it is less likely in group 1, and OR=1\text{OR} = 1 means no association.

Key Formula

OR=a/bc/d=adbc\text{OR} = \frac{a/b}{c/d} = \frac{ad}{bc}
Where:
  • aa = Count in group 1 with the outcome
  • bb = Count in group 1 without the outcome
  • cc = Count in group 2 with the outcome
  • dd = Count in group 2 without the outcome
  • OR\text{OR} = Odds ratio comparing group 1 to group 2

Worked Example

Problem: In a drug trial, 50 patients get a new treatment and 50 get a placebo. In the treatment group, 40 recover and 10 do not. In the placebo group, 20 recover and 30 do not. Find the odds ratio of recovery for treatment compared with placebo.
Step 1: Find the odds of recovery in the treatment group. Divide recoveries by non-recoveries.
oddstreatment=4010=4\text{odds}_{\text{treatment}} = \frac{40}{10} = 4
Step 2: Find the odds of recovery in the placebo group.
oddsplacebo=2030=23\text{odds}_{\text{placebo}} = \frac{20}{30} = \frac{2}{3}
Step 3: Divide the treatment odds by the placebo odds. The cross-product shortcut adbc\frac{ad}{bc} gives the same result.
OR=42/3=40⋅3010⋅20=6\text{OR} = \frac{4}{2/3} = \frac{40 \cdot 30}{10 \cdot 20} = 6
Answer: The odds ratio is 6. The odds of recovering are 6 times higher with the treatment than with the placebo.

Why It Matters

Odds ratios are the standard effect size in medical case-control studies and in logistic regression, which you will meet in college statistics and epidemiology courses. Researchers usually report an odds ratio together with a confidence interval. If that interval contains 1, the data do not show a significant association at that confidence level.

Common Mistakes

Mistake: Reading an odds ratio as a ratio of probabilities (relative risk), for example saying the treatment makes recovery '6 times as likely' in the example above.
Correction: Odds are not probabilities. In the example, the recovery probabilities are 4050=0.8\frac{40}{50} = 0.8 and 2050=0.4\frac{20}{50} = 0.4, so the relative risk is 22, even though the odds ratio is 66. Say 'the odds are 6 times higher,' not 'the chance is 6 times higher.' The two measures are close only when the outcome is rare.

Related Terms

  • Ratio — An odds ratio is a ratio of two odds
  • Odds — The odds that an odds ratio compares
  • Confidence Interval — Gives a plausible range for the true odds ratio
  • Hypothesis Testing — Tests whether the odds ratio differs from 1
  • Null Hypothesis — Usually states that the odds ratio equals 1
  • Standard Error — Measures the sampling variability of the log odds ratio
  • p-value — Measures the evidence against an odds ratio of 1
  • Probability — Odds are built from probabilities of outcomes