13 Research methods, statistics, critical review and evidence-based practice13.2 Epidemiological studies
13.2.2.3 Prevalence, cumulative incidence and incidence rate
4 min read2 figures
Written by Resident Doctor Committee · Published
The three measures
Measure | Numerator | Denominator | What it describes |
|---|---|---|---|
Prevalence | All existing cases, old and new | Whole population, cases included | Current burden |
Cumulative incidence | New cases in a fixed period | Disease-free people at risk at the start | Risk over the period |
Incidence rate | New cases | Total person-time at risk | New cases per unit of person-time |
Incidence counts only new cases arising in a defined period, so a report of new cases over a time window is an incidence statement, not a prevalence one.
A cross-sectional survey is a single snapshot with no follow-up, so it measures prevalence only.
A cross-sectional survey cannot give incidence, cumulative incidence, incidence rate or incidence-based measures of association, because these need a disease-free population followed over time.
Prevalence
Type | Who is counted |
|---|---|
Point prevalence | All existing cases, old and new, at a single moment |
Period prevalence | Anyone with the condition at any time in a defined interval: cases present at the start plus new cases arising during it, usually over the mid-period population |
Lifetime prevalence | People alive on the survey day who have ever had the disorder |
Prevalence sizes the caseload a service must plan for, whereas incidence counts only new cases and underestimates demand, especially in chronic disorders.
Prevalence and screening-test studies
A diagnostic-accuracy study, which validates a screening test, gives population prevalence only if its sample is known to represent the target population; a convenience sample of clinic attenders does not.
Negative predictive value is the proportion of negative results that are true negatives, so it tells a person with a negative result how likely they are to be free of the condition.
Cumulative incidence
Cumulative incidence (incidence proportion, risk) is new cases during a fixed period divided by the disease-free people at risk at its start. It is a proportion that estimates an individual's risk.
People who already have the condition at the start are excluded from the denominator, and only new cases enter the numerator.
Risk has no units and lies between 0 and 1, but it needs a stated time period: 15% over six months differs from 15% over five years.
Cumulative incidence suits closed (fixed) populations followed for a set period, because it assumes everyone is followed for the whole period.
Attack rate is all cases arising after a single acute exposure divided by all people exposed at the outset, without netting out later deaths.
Comparing incidence between two periods needs a population-at-risk denominator for each period, because a count of new cases is only the numerator.
Incidence rate
The incidence rate (incidence density) is new cases divided by total person-time at risk, the summed disease-free observation time of every participant.
Each person contributes time only while observed and at risk, and leaves the risk set at the event, at loss to follow-up or at censoring.
Person-time copes with unequal follow-up, staggered entry, losses, censoring and competing events such as emigration or death from other causes.
The incidence rate suits dynamic populations, which gain and lose people at risk; a mid-year population can approximate annual person-years.
A rate has units of inverse time, such as per person-year, and can exceed 1.
Worked example
Ten patients in remission after first-episode psychosis are followed for up to 24 months. Four relapse, two are lost to follow-up, and together they give 14.5 person-years at risk.
Measure | Calculation | Result |
|---|---|---|
Point prevalence at month 12, when two are in relapse | 2 / 10 | 20% |
Cumulative incidence over 24 months | 4 / 10 | 0.40 |
Incidence rate | 4 / 14.5 person-years | 0.276 per person-year, or 27.6 per 100 person-years |
Lifetime morbid risk
Lifetime morbid risk is the probability of developing the disorder over the whole lifespan, estimated on the entire birth cohort, including the dead and those not yet through the age of risk.
Lifetime prevalence differs because it counts only people alive on the survey day who have ever had the disorder.
Prevalence, incidence and duration
In a steady state, prevalence is roughly incidence multiplied by mean duration, so mean duration = prevalence / incidence.
A disease-specific death rate given alongside does not enter the calculation.
The approximation needs stable incidence, duration and population flows, and holds best when prevalence is low.
Schizophrenia incidence of 0.00015 per person-year (15 per 100,000) multiplied by a mean duration of 25 years predicts a prevalence of 0.00375, about 3.8 per 1,000.
Halving duration to 12.5 years halves predicted prevalence to about 1.9 per 1,000, with incidence unchanged.
Change | Effect on prevalence |
|---|---|
Cure or treatment that acts only after onset | Duration shortens, so prevalence falls while incidence is unchanged |
Lower mortality without cure | People live longer with the disease, so prevalence rises |
Death, reduced detection or prevention | Prevalence falls |
When incidence is equal in two groups, or when their prevalence ratio differs from their incidence ratio, the difference in prevalence is best explained by a difference in duration of illness.
A rare condition with short episodes, ending in death or recovery, has low incidence and low prevalence.