The incidence rate calculator above applies the standard epidemiological definition: new cases divided by the person-time at risk. It reports the rate on whichever per-person-time scale you choose, the incidence proportion for the same cohort, the person-time denominator it used, and an approximate confidence interval so the precision of the estimate is visible alongside the estimate itself.
Arb Digital publishes this as an explanation of a definition. It attaches no verdict to any rate. A number produced here is not high, low, alarming or reassuring; it is an estimate whose meaning depends entirely on the population, the case definition, the follow-up and the comparison you intend to make. Where the question is about comparing two groups rather than describing one, our relative risk calculator works from a two-by-two table of counts and is the right tool.
What This Incidence Rate Calculator Does
It computes two related but distinct quantities that are frequently confused. The incidence rate is new cases divided by person-time, and it has units of one over time — cases per person-year. It can in principle exceed one and is not a probability. The incidence proportion, also called cumulative incidence or risk, is new cases divided by the number of people at risk at the start. It is dimensionless, runs from zero to one, and is only meaningful when the follow-up period is stated alongside it.
Person-time is the part that requires a decision, so the tool makes it an explicit choice rather than an assumption. In a closed cohort with no losses and no cases, person-time is simply the population multiplied by the follow-up period. Once people become cases they stop contributing time at risk, so the mid-point convention subtracts half a period for each case, on the assumption that cases arise uniformly through the window. Where you have individual records, summing each participant's own time is exact and is always preferable.
The confidence interval uses the standard normal approximation on the log of the rate, which is reasonable when the case count is moderate. On very small counts that approximation is poor, and the tool says so rather than presenting an interval that looks more trustworthy than it is.
How to Use It
- Count only new cases. Anyone who already had the condition at the start is neither a case nor part of the population at risk.
- Enter the population at risk, meaning people who could become a case. Those already immune, already affected or otherwise not susceptible are excluded.
- Choose the person-time convention deliberately, and record which one you used. It changes the denominator and therefore the rate.
- Pick the reporting multiplier that matches whatever you are comparing against, since per 1,000 and per 100,000 differ by a factor of a hundred.
- Read the confidence interval before quoting the point estimate. On small counts the interval is what carries the honest information.
The Formula and How It Is Calculated
The incidence rate is IR = new cases ÷ total person-time at risk. The incidence proportion is IP = new cases ÷ population at risk at the start. The confidence interval used here is the log-transformed interval, with the rate multiplied and divided by the exponential of 1.96 divided by the square root of the case count.
Work the default through by hand. There are 45 new cases among 5,000 people at risk followed for 4 years. Under the closed-cohort convention person-time is 5,000 × 4 = 20,000 person-years, so the rate is 45 ÷ 20,000 = 0.00225 per person-year, which is 225.0 per 100,000 person-years. The incidence proportion is 45 ÷ 5,000 = 0.009, or 0.9 percent over four years — and that period must be quoted with it, because a proportion without a time window is meaningless.
Under the mid-point convention the 45 cases each contribute half the period rather than all of it, so person-time is (5,000 − 22.5) × 4 = 19,910 person-years and the rate becomes 45 ÷ 19,910 = 0.0022602, or 226.0 per 100,000 person-years. The difference here is small because the case count is small relative to the cohort; in a study where a large fraction becomes a case it is substantial. The CDC's Principles of Epidemiology teaching materials set out the same definitions and the person-time convention in detail.
Rate, Proportion and Prevalence Are Three Different Things
These three get used interchangeably in ordinary writing and mean genuinely different things, and mixing them produces conclusions that do not follow.
Incidence rate counts new cases per unit of person-time. It measures how fast new cases are appearing. Its units are per time, so it has no upper bound and is not a probability. Incidence proportion counts new cases per person followed, over a stated period. It is a probability, bounded by one, and it always requires its time window to be quoted. Prevalence counts existing cases at a point in time, divided by the whole population. It includes cases that arose years ago and says nothing about how fast new ones appear.
The relationship between them is the reason confusing them matters. Prevalence depends on both incidence and duration: a condition with low incidence but a very long duration can have high prevalence, and a condition with high incidence but rapid resolution can have low prevalence. Reading a high prevalence as evidence of rapidly rising incidence, or a falling prevalence as evidence of falling incidence, are both errors, and both are common. Prevalence can also fall because cases resolve faster or because people die sooner, neither of which is an improvement in incidence.
Why Crude Rates Cannot Be Compared Directly
This is the most important caution on this page. A crude incidence rate reflects both how frequently the condition arises and what the population looks like, and for most conditions the second effect is large.
Almost every disease has a strong age gradient. A population with a higher proportion of older people will show a higher crude rate for an age-associated condition even if the risk at every single age is identical to a younger population's. Comparing two crude rates therefore mixes a real difference in risk with an artefact of the age structures, and there is no way to separate them from the crude figures alone.
The remedy is standardisation. Direct standardisation applies both populations' age-specific rates to a single reference age structure, producing rates that are comparable by construction. Indirect standardisation applies a reference set of rates to each population's own structure. Both require age-specific case and population counts, which is more data than this calculator takes, and neither can be reconstructed from a crude rate afterwards. The practical rule: a crude rate describes one population, and comparing crude rates across populations, regions or time periods without standardisation is not a valid comparison. The World Health Organization's indicator metadata registry documents the exact definitions and standardisation choices behind published health indicators.
Where Person-Time Goes Wrong
Person-time is the concept that makes incidence rates work and it is also where most errors enter. Three cases are worth naming.
Loss to follow-up. Someone who leaves the study after a year contributed one person-year, not four. Treating them as if they were followed throughout inflates the denominator and understates the rate. If people leave for reasons related to the outcome, the resulting bias is not merely a matter of size and cannot be fixed by arithmetic.
Late entry. Participants who join partway through contribute only from their entry date. Open cohorts where people enter and leave continuously need individual time accumulated properly, which is exactly what the exact person-time option is for.
Time after the event. Once someone becomes a case they are no longer at risk of becoming a first case, so their time stops. For recurrent events the definition changes, and you must state whether you are counting first events or all events, because the two give different numbers from the same data. If you are working with counts of rare events, our Poisson distribution calculator gives the underlying probability model, the confidence interval calculator handles intervals for other statistics, and the sample size calculator works out how much follow-up would be needed for a given precision.
Arb Digital builds free tools that report the assumption behind every figure rather than burying it.
Browse All Free Tools Talk To Our TeamCommon Mistakes to Avoid
- Including existing cases — they belong to prevalence, and counting them as new cases inflates the numerator while wrongly enlarging the population at risk.
- Quoting an incidence proportion without its time period — a risk of one percent means nothing until you say one percent over what length of follow-up.
- Comparing crude rates across populations — differences in age structure alone can produce large differences in crude rates with no difference in underlying risk.
- Assuming everyone was followed for the full period — losses, late entries and the time after a case all reduce person-time, and ignoring them understates the rate.
- Reading a wide confidence interval as a finding — it means the estimate is imprecise because the counts were small, not that the rate is high or low.
Related Free Tools From Arb Digital
Compare two groups with the relative risk calculator, model rare event counts with the Poisson distribution calculator, build intervals for other statistics with the confidence interval calculator, plan a study with the sample size calculator, or work through death rates with the mortality rate calculator. The full free online tools hub lists every statistics tool we publish.
Frequently Asked Questions
It is the number of new cases divided by the total person-time at risk. Its units are cases per unit of time, so it measures how fast new cases appear rather than the chance that any one person becomes a case.
The proportion divides new cases by the number of people at risk at the start, giving a dimensionless probability between zero and one that must always be quoted with its follow-up period. The rate divides by person-time instead.
The sum of the time each participant spent at risk. Time stops when someone becomes a case, is lost to follow-up, or reaches the end of the study, and starts when they enter.
Because they give different denominators from the same data. A closed cohort with no losses can use population times period; where cases accumulate, the mid-point adjustment is closer; individual records are exact.
Prevalence counts existing cases at a point in time and reflects both how often cases arise and how long they last. A condition with low incidence and long duration can have high prevalence.
Not without checking that the case definition, the population at risk, the person-time convention and the standardisation all match. Crude rates from different populations are not comparable without standardisation.
How precisely the rate is estimated given the number of cases observed. A wide interval reflects few cases and says nothing about whether the rate itself is high or low.
Yes. It is a rate per unit of time rather than a probability, so it has no upper bound of one. An incidence proportion cannot exceed one.
This page explains a standard epidemiological definition for educational purposes and is not medical advice. It attaches no interpretation to any rate. For guidance on health data or any personal health question, consult a qualified clinician or your national public health authority.
