Skip to content
All articles

Statistics

Confidence Intervals Explained: How to Read and Report Them Correctly

A confidence interval (CI) gives a range of plausible values for the true population value, based on your sample. A 95% CI means that if the study were repeated many times, about 95% of the intervals built this way would contain the true value

Resora Research Team · 14 Nov 2026 · 5 min read

A confidence interval (CI) gives a range of plausible values for the true population value, based on your sample. A 95% CI means that if the study were repeated many times, about 95% of the intervals built this way would contain the true value. It is calculated as the estimate ± 1.96 × standard error. A CI shows both the size of an effect and its precision: a narrow interval means a precise estimate, usually from a large sample. For a difference between groups, a CI that excludes 0 means p < 0.05; for a ratio such as an odds ratio or relative risk, a CI that excludes 1 means p < 0.05. Always report effect sizes with 95% CIs, because they carry more information than a p-value alone.

Why a single number is not enough

Suppose a survey of 250 adults finds that 32% are hypertensive. That 32% is only an estimate: a different random sample of 250 would give 29% or 35%. The true population figure is unknown. A confidence interval tells you how far the truth could plausibly be from your estimate.

The general form is:

\text{CI} = \text{estimate} \pm (\text{critical value} \times SE)
Confidence levelCritical value (large samples)
90%1.64
95%1.96
99%2.58

For small samples of continuous data, use the t value for your degrees of freedom instead of 1.96 (for n = 10, t = 2.26).

CI for a mean

95\%\ CI = \bar{x} \pm 1.96 \times \frac{SD}{\sqrt{n}}

Worked example. Mean systolic BP in 50 patients is 132 mmHg with an SD of 12.

  1. SE = 12 / √50 = 1.70 mmHg
  2. Margin = 1.96 × 1.70 = 3.3
  3. 95% CI = 132 ± 3.3 = 128.7 to 135.3 mmHg

Report as: mean SBP 132 mmHg (95% CI 128.7 to 135.3).

CI for a proportion

95\%\ CI = p \pm 1.96 \times \sqrt{\frac{p(1-p)}{n}}

Worked example. 80 of 250 adults are hypertensive, so p = 0.32.

  1. SE = √(0.32 × 0.68 / 250) = 0.0295
  2. Margin = 1.96 × 0.0295 = 0.058
  3. 95% CI = 0.32 ± 0.058 = 26.2% to 37.8%

Report as: prevalence 32% (95% CI 26.2% to 37.8%). This formula works well when np and n(1 − p) are both at least 5; for very small samples or proportions near 0% or 100%, use the exact (Clopper–Pearson) or Wilson method.

CI for a difference or a ratio

For a difference between two means or proportions, the 95% CI is the difference ± 1.96 × the SE of the difference. For a ratio (odds ratio, relative risk, hazard ratio), the CI is calculated on the log scale and then converted back, which is why it is not symmetric around the estimate. Statistical software does this for you. What matters is how to read the result.

How to interpret a confidence interval

Correct reading: "We are 95% confident that the true prevalence lies between 26.2% and 37.8%." More precisely, the method used captures the true value in 95% of repeated samples.

Common wrong reading: "There is a 95% probability that the true value lies in this particular interval." Once calculated, the interval either contains the true value or it does not.

What makes a CI narrower

FactorEffect on width
Larger sample sizeNarrower (four times the sample halves the width)
Less variability (smaller SD)Narrower
Lower confidence level (90% instead of 95%)Narrower, but less certain
Higher confidence level (99%)Wider

A wide interval signals an imprecise estimate, usually because the study is small.

Using the CI to judge significance

A CI contains the same information as the p-value, plus more.

Comparing two groupsThe 95% CI is…Meaning
Difference in means or proportionsDoes not include 0p < 0.05, statistically significant
Difference in means or proportionsIncludes 0p > 0.05, not significant
Odds ratio, relative risk or hazard ratioDoes not include 1p < 0.05, statistically significant
Odds ratio, relative risk or hazard ratioIncludes 1p > 0.05, not significant

Example. An odds ratio of 2.7 with a 95% CI of 1.4 to 5.0 is significant (the interval excludes 1) and shows the effect could be anything from modest to large. An odds ratio of 2.7 with a CI of 0.8 to 9.1 is not significant, and the wide interval warns that the study was too small.

Statistical versus clinical significance

The CI also helps judge importance. If a new drug lowers BP by 3 mmHg (95% CI 1 to 5), the result is significant, but even the upper limit may fall short of what clinicians consider worthwhile. If a drug lowers BP by 8 mmHg (95% CI −1 to 17), the result is not significant, yet the data are compatible with a large benefit. A study this size is inconclusive, not negative.

Common mistakes

  1. Reporting only the p-value and leaving out the CI.
  2. Reading "not significant" as "no effect" when the CI is wide.
  3. Mixing up the CI of the mean with the range in which individual patients fall. The CI of the mean is narrow; the range containing 95% of individuals (mean ± 1.96 SD) is much wider.
  4. Concluding that two groups differ because their separate CIs do not overlap, or that they are the same because they do. Compare them directly by calculating the CI of the difference.
  5. Using the simple proportion formula for small samples or extreme percentages.
  6. Not stating the confidence level. Say "95% CI" every time.

How to report

  • Write the estimate first, then the interval in brackets: "OR 2.7 (95% CI 1.4 to 5.0)".
  • Use "to" rather than a dash when numbers can be negative.
  • Report the exact p-value as well, but treat the CI as the main result.

Keywords

confidence interval, 95% confidence interval, how to calculate confidence interval, confidence interval formula, confidence interval interpretation, CI for mean, CI for proportion, confidence interval vs p-value, margin of error, standard error, confidence interval odds ratio, statistical significance, precision of estimate, how to report confidence interval, biostatistics for thesis

References

  • Gardner MJ, Altman DG. Confidence intervals rather than P values: estimation rather than hypothesis testing. BMJ. 1986;292:746–750.
  • Altman DG, Machin D, Bryant TN, Gardner MJ. Statistics with Confidence. 2nd ed. BMJ Books; 2000.
  • Kirkwood BR, Sterne JAC. Essential Medical Statistics. 2nd ed. Blackwell Science; 2003.
  • Greenland S, et al. Statistical tests, P values, confidence intervals, and power: a guide to misinterpretations. European Journal of Epidemiology. 2016;31:337–350.

Confidence Intervals Explained: How to Read and Report Them Correctly · Resora