To translate incidence and prevalence accurately, a translator has to preserve who is counted, when they are counted and whether the numerator contains new cases or all existing cases. Incidence usually concerns new events occurring during a period. Prevalence concerns people or units that have the condition at a point or during a period, including cases that began earlier. Both can be written as percentages or rates, and both can involve the same condition, which is why a seemingly harmless synonym such as “frequency” can erase an important epidemiological distinction.
This guide explains how to translate incidence, cumulative incidence, incidence rate, person-time, point prevalence, period prevalence, attack rate and related population-frequency language in public-health reports, epidemiological studies, research articles, technical dashboards and educational material. The search intent is practical: how do you translate a disease or event frequency without swapping new cases for existing cases, people for person-years, a one-day snapshot for a one-year period, or a proportion for a rate? The answer is to label numerator, denominator and time window before rewriting the sentence.
A reliable translation also preserves the population at risk. “Twenty cases per 100,000 people per year” is not the same statement as “twenty percent per year.” “Ten new cases among 1,000 people followed for a year” is not necessarily interchangeable with “ten cases per 1,000 person-years” when follow-up differs. The translator should therefore treat time, population and case status as part of the metric itself rather than as background wording that can be shortened freely.
A fifty-second orientation
CDC epidemiology guidance distinguishes incidence from prevalence by the numerator. Incidence counts new cases during a specified period. Prevalence counts all cases present at a point or during a period, regardless of when they began. An incidence proportion uses a population at risk as its denominator. An incidence rate, also called a person-time rate in many contexts, uses accumulated time at risk in the denominator.
Those distinctions immediately create a translation rule: do not translate every epidemiological percentage or “rate” with one generic noun. First identify whether the quantity is a proportion, a rate, a count or a prevalence estimate. Then preserve its time frame and the population that contributed to the denominator.
1. Incidence starts with new events
Incidence focuses on occurrence. A person or unit enters the numerator when a new event begins during the defined observation period. If the condition was already present before the period started, that existing case usually does not belong in an incidence numerator for first occurrence under the stated definition.
Translate new, newly diagnosed, onset, first occurrence and incident case carefully. A source may define incident case by symptom onset, first positive test, first recorded diagnosis or another operational date. Do not substitute a more familiar event date unless the study defines them as equivalent.
A useful working note is: “numerator = new cases meeting definition during period.” That one line prevents a translator from later replacing incidence with “number of people with the condition,” which would include pre-existing cases and move the concept toward prevalence.
2. Prevalence starts with current or period-present cases
Prevalence asks how common an existing condition is in a population at a specified time or over a defined period. The numerator includes cases that may have begun before the observation window and are still present under the study definition.
Translate “existing cases,” “current cases” and “persons with the condition” without accidentally implying new onset. A prevalence sentence does not necessarily tell the reader when the condition started. It describes presence in the population under the specified time frame.
When a source compares incidence and prevalence in the same paragraph, resist stylistic variation that makes the terms sound interchangeable. Repetition can be useful. It is better to repeat the correct technical noun than to replace one occurrence with a loose synonym that changes the numerator.
3. Point prevalence is a snapshot
Point prevalence refers to the proportion of a population with the condition at a specified point in time. A source may use a particular date, survey day, census date or measurement visit as the snapshot.
If 50 of 1,000 people have a condition on the survey date, point prevalence is 5% under that simple definition. The statement does not tell you how many new cases developed during the month or how long those cases will continue.
Preserve the snapshot date or phrase such as “on 1 July,” “at baseline” or “at the survey visit.” Removing that reference can make point prevalence sound like a timeless property of the population.
4. Period prevalence covers a window
Period prevalence concerns people who had the condition at any relevant time during a specified period, depending on the source definition. It can include cases present at the start and cases developing during the interval.
Translate “during the previous twelve months,” “at any time in the quarter” or another period qualifier with the prevalence metric. A one-year period prevalence is not the same as point prevalence measured on the final day of that year.
Do not infer the exact calculation from the phrase alone if the source uses a specialised denominator or repeated survey design. Preserve the study’s method. The general concept tells you what to look for, but the operational definition controls the final translation.
5. Incidence proportion is a risk-like proportion over a period
Incidence proportion, sometimes called cumulative incidence or risk in an appropriate context, uses new cases in the numerator and an initially at-risk population in the denominator. It is tied to a specified observation period.
Suppose 20 of 1,000 initially disease-free participants develop the condition over one year with complete follow-up in this simplified example. The one-year incidence proportion is 2%. The period matters: a five-year incidence proportion would answer a different question even in the same cohort.
Translate “risk” cautiously because the everyday word can imply causation, danger or an individual prediction. If the source formally defines risk as cumulative incidence, preserve the statistical meaning. Do not automatically replace every incidence proportion with “chance” in technical writing.
6. Incidence rate uses person-time
An incidence rate uses accumulated time at risk in the denominator. Participants can contribute different amounts of time because they enter late, leave early, experience the event, die, move away or reach the end of follow-up at different moments.
If ten new cases occur during 5,000 person-years of observation, the incidence rate is 2 cases per 1,000 person-years. This is a rate, not a statement that two people out of every thousand will necessarily develop the condition in one calendar year.
Preserve person-time units exactly. “Per 1,000 person-years” should not become “per 1,000 people each year” unless the design and interpretation justify that simplification. Person-time combines people and observation duration in one denominator.
7. Person-years are accumulated observation time, not a headcount
One hundred person-years can arise from one hundred people observed for one year each, fifty people observed for two years each, or many unequal contributions that sum to the same total. The denominator does not uniquely reveal how many people were enrolled.
Translate person-years, person-months, participant-days and similar compound units as time-at-risk measures. Do not shorten person-years to years. “Five events per 1,000 years” removes the people from the denominator and becomes meaningless in most epidemiological contexts.
Likewise, do not infer that one person can contribute unlimited at-risk time after experiencing a first-event outcome. The study rules determine when follow-up stops or whether recurrent events are counted. Translate the protocol’s definition rather than applying a generic rule.
8. Rate and proportion are not interchangeable
A proportion has a numerator that is part of the denominator and typically ranges from zero to one, or zero to one hundred percent. A person-time rate has events divided by time at risk and is not bounded by one in the same way, especially when recurrent events are possible.
Some public-health writing uses “rate” loosely for proportions, such as an attack rate that is mathematically a proportion. Translators should follow the established source term while understanding the underlying quantity. Do not “correct” a conventional label without authority.
A glossary can store both the conventional name and mathematical type. For example: “attack rate — conventional epidemiological term; usually a proportion over an outbreak period.” This helps preserve domain vocabulary while preventing conceptual confusion elsewhere.
9. Attack rate is usually outbreak-period incidence proportion
In outbreak investigations, attack rate often means the proportion of an at-risk population that becomes ill during a defined outbreak period. It is commonly expressed as a percentage even though the traditional name uses rate.
Suppose 30 of 150 attendees at a fictional event meet the case definition. The attack rate is 20% under that definition. The denominator should include the relevant at-risk attendees, not everyone in the city or every person contacted later.
Translate the event, exposure group and case definition where the source provides them. Attack rate can be compared across foods, rooms or groups in an investigation, and the denominator may differ for each exposure-specific calculation.
10. Secondary attack rate has a narrower denominator
Secondary attack rate is often used to describe spread among susceptible contacts of primary cases within a defined group or setting. The denominator typically excludes the primary cases and may exclude people not susceptible or not exposed, depending on the study definition.
Do not translate it as simply “second attack rate.” The word secondary describes transmission generation or contact context, not the second time someone calculated a percentage.
If the source provides a household secondary attack rate, preserve household and the contact denominator. A general “infection rate” may lose the specific relationship being measured.
11. Mortality rate and case-fatality proportion answer different questions
Mortality rate usually relates deaths to a population or person-time denominator. Case-fatality proportion relates deaths among identified cases to the number of cases. The same number of deaths can therefore produce very different percentages depending on which population is used.
Suppose a fictional population of 100,000 has 1,000 cases and 20 deaths among those cases. The case-fatality proportion is 2%. The population mortality proportion for the period would be 20/100,000, or 0.02%, if that denominator and period are appropriate. These are not interchangeable statements.
Translate fatality, mortality and death rate with care. A phrase such as “2% mortality” can be ambiguous without context. Preserve the source denominator or make it explicit when the text already defines it.
12. Numerator case definitions control comparability
An incidence or prevalence number depends on what counts as a case. The source may require symptoms plus a laboratory test, a diagnosis code, self-report, a threshold score or another operational definition.
If the case definition changes between years, apparent changes in incidence or prevalence may partly reflect the new definition. A translation should preserve notes such as “using the revised case definition” or “based on laboratory-confirmed cases only.”
Do not strengthen “probable cases” into “confirmed cases” for simplicity. Probable, suspected, confirmed and possible can represent distinct surveillance categories. Their differences directly affect numerators.
13. Denominator population definitions matter just as much
Rates and proportions are interpretable only when the denominator population is known. Residents, registered patients, survey respondents, workers on site, people at risk and total population are not necessarily the same group.
A source may use mid-year population estimates, census counts, average population, people under surveillance or person-time. Translate the denominator label rather than replacing it with the generic word population.
When comparing two regions, check whether both use the same population basis. A target sentence that drops “age 65 and older” from one rate but not the other creates a misleading comparison even if both numerical rates are copied correctly.
14. “Per 100,000” is scaling, not a new denominator population
A rate reported as 25 per 100,000 may be a scaled version of an underlying proportion or rate so readers can compare small frequencies. It does not mean the study literally observed exactly 100,000 people.
Keep the scaling factor with the unit. “25 cases per 100,000 person-years” and “25 cases per 100,000 population” are different. The first is a person-time rate; the second may be a population-based measure tied to a specified period.
Do not convert 25 per 100,000 to 0.025% unless the quantity is truly a proportion and the target context benefits from conversion. A person-time rate cannot be converted to a percentage merely by moving the decimal point.
15. Crude and adjusted rates should remain distinct
A crude rate reflects the observed frequency in the population without certain standardisation adjustments. An age-adjusted or standardised rate reweights data to a reference population or method so comparisons can be less affected by differing age structures.
Translate age-adjusted, age-standardised, directly standardised and indirectly standardised according to the source method. Do not shorten them all to “corrected rate.” The adjustment has a specific statistical meaning.
Adjusted rates may not correspond to the actual proportion observed in the local population. A target sentence should not say “X people out of every 100,000 experienced the event” when X is a standardised rate intended for comparison rather than a literal observed proportion.
16. Standardised rates depend on a reference population
When a report uses age standardisation, the choice of standard population affects the reported adjusted rate. Two publications can therefore produce different standardised values from the same underlying age-specific rates if they use different standards.
Preserve the named standard population and version where the source gives it. “World standard population,” “national standard population” or a specific census standard should not disappear from a methods note if comparability depends on it.
If the source compares crude and standardised trends, keep the labels in figure legends and tables. A chart with two lines can become misleading if both are translated simply as “rate.”
17. Age-specific and sex-specific rates are subgroup measures
An age-specific rate uses the relevant age group in both numerator and denominator. A sex-specific rate uses the corresponding subgroup. These are not adjustments; they are direct rates within defined strata.
Translate the subgroup label as part of the metric: “incidence among adults aged 65–74,” not merely “incidence.” If the graph uses abbreviated strata, ensure the legend survives responsive layouts or table exports.
A target summary should not generalise the highest subgroup rate to the whole population. Preserve whether the source describes a subgroup, total population or standardised comparison.
18. Incidence density can be another name for a person-time rate
Some epidemiological literature uses incidence density for an incidence rate based on person-time. Terminology varies across disciplines and periods. Translate the source term consistently and include the mathematical definition where ambiguity is possible.
Do not infer that density means geographic density. In this context it describes event occurrence relative to accumulated observation time. A literal translation of density into a spatial concept can mislead readers.
A bilingual glossary should therefore include incidence rate, person-time rate and incidence density as potentially related terms while noting the project’s preferred label and source definition.
19. Recurrent events need an explicit counting rule
Some studies count only the first event per person. Others count repeated events. If a person can contribute multiple events to an incidence rate, the numerator may exceed the number of people who experienced at least one event.
Translate events, episodes, people and cases separately. “Thirty episodes among twenty people” should not become “thirty patients.” The distinction can matter greatly in infection recurrence, hospitalisation, injuries, machine failures or other repeat-event settings.
If the source reports recurrent-event rate per person-time, preserve the event-counting rule. Do not compare it directly with a first-event cumulative incidence without acknowledging that the numerators are constructed differently.
20. Open populations require careful denominator language
In an open or dynamic population, people can enter and leave over time. A period population count may therefore be approximated using mid-period population, average population or person-time, depending on the method.
Translate resident population, average population, mid-year population and population at risk as the source specifies. Do not assume a mid-year denominator is a literal count of everyone ever observed during the year.
When a rate is based on administrative population estimates, preserve the source agency and estimate year if they are part of the methods. This helps readers understand how the denominator was constructed.
21. Changes in testing can change measured incidence
Observed incidence can rise when more cases are detected because testing expands, surveillance improves or reporting rules change, even if the underlying occurrence changes differently. A translation should preserve the author’s distinction between recorded incidence and underlying disease occurrence when the source makes it.
Words such as reported cases, detected cases, notified cases and estimated infections are not interchangeable. Each can refer to a different surveillance layer. Do not streamline them into one term throughout a report.
Likewise, a decline in recorded incidence after a reporting-system change should not automatically be translated as “fewer people became ill” unless the source supports that causal interpretation.
22. Prevalence depends on incidence and duration, but not by a universal shortcut
In a stable simplified setting, prevalence can be related to incidence and average duration of disease. This relationship is often taught conceptually: conditions with long duration can have high prevalence even when new cases are relatively uncommon.
Do not use a simple prevalence ≈ incidence × duration relationship as an automatic conversion rule in every real population. The approximation depends on assumptions about stability, rarity, population flow and disease dynamics.
Translate conceptual explanations as explanations, not as exact identities unless the source specifies the assumptions. A long-duration chronic condition and a short-duration acute condition can have very different prevalence for the same incidence.
23. Worked case: incidence proportion versus incidence rate
Consider a fictional cohort of 2,000 initially at-risk people followed for varying periods. During the study, 40 first events occur. If complete follow-up for a fixed one-year period applied to everyone, an incidence proportion could be 40/2,000 = 2% over one year. But suppose actual observation totals 1,500 person-years because participants entered late or left early.
The person-time incidence rate is then 40/1,500 = 0.0267 events per person-year, or about 26.7 per 1,000 person-years. The 2% proportion and 26.7-per-1,000-person-years rate answer related but different questions and rely on different denominator assumptions.
A flawed translation might convert the rate to “2.67% per year.” That simple conversion is not generally justified because a person-time rate is not automatically a one-year risk. The repaired target retains the person-time unit and describes the incidence proportion separately if the source reports it.
The teaching example shows why the denominator must travel with the number. If a target table drops “person-years” from the column heading, the displayed values can no longer be interpreted correctly.
24. Worked case: point prevalence and annual incidence
Imagine a fictional town of 10,000 people. On 1 January, 500 residents are living with Condition X, giving point prevalence of 5% on that date. During the following year, 100 new cases are identified among the relevant at-risk population under the study definition.
A target text should not say “the annual incidence was 5%” because 5% describes prevalence at the starting snapshot. Nor should it say “100 people had the condition” because 500 already had it at the point-prevalence date and new cases later occurred.
The incidence calculation depends on the at-risk denominator and follow-up rules, which this simplified example has not fully specified. Therefore the translator can preserve the count of 100 new cases without inventing an incidence percentage. Knowing when not to calculate is part of accurate technical translation.
A source may intentionally present prevalence and incidence side by side to show burden and new occurrence. Keep both because they answer complementary questions.
25. Worked case: crude and age-standardised rates
Suppose fictional Region A has a crude mortality rate of 800 per 100,000 and an age-standardised rate of 500 per 100,000 using a named standard population. Region B has a crude rate of 600 and an age-standardised rate of 520 under the same standard.
The crude comparison suggests a higher observed population rate in A. The age-standardised comparison is much closer and slightly higher in B. A translation that labels both sets simply “mortality rate” removes the reason the table contains two rows.
Preserve crude and age-standardised in headings, captions and commentary. Do not interpret the standardised value as the literal number of deaths that occurred per 100,000 residents. It is a comparative statistical construct under the specified standardisation method.
If the source explains that age structure accounts for part of the crude difference, translate that explanation. Do not invent a causal reason beyond what the standardisation supports.
26. Practice clinic with explained answers
Practice one: incidence numerator. Fifty people already have a condition at baseline and ten new cases develop during follow-up. For first occurrence, the incidence numerator is the ten new cases, not sixty. Baseline cases may contribute to prevalence instead.
Practice two: point prevalence. At a survey date, 80 of 2,000 people have the condition. Point prevalence is 4%. This does not tell you how many of those 80 were newly diagnosed that day.
Practice three: person-time. Twelve events occur over 3,000 person-years. The rate is four per 1,000 person-years. Do not translate it as four per 1,000 people without the time unit.
Practice four: attack rate. Twenty-five cases occur among 100 attendees under the outbreak case definition. The attack rate is 25%. The conventional name rate does not make the denominator person-time.
Practice five: case fatality. Ten deaths occur among 500 cases. The case-fatality proportion is 2%. This is not a population mortality rate unless the case count equals the whole population, which it usually does not.
Practice six: scaling. A proportion of 0.0004 can be written as 40 per 100,000 if the measure is a compatible proportion. Do not apply the same conversion to a person-time rate without retaining its time unit.
Practice seven: standardisation. An age-standardised rate should retain its reference population. Do not describe it as the observed crude rate in the actual population.
Practice eight: period prevalence. “Had the condition at any time during the year” is broader than “had the condition on 31 December.” Translate the period qualifier explicitly.
Practice nine: recurrent events. Thirty hospitalisations among twenty people represent thirty events but twenty people with at least one event. Preserve which numerator the source reports.
Practice ten: open population. A rate based on mid-year population should retain that denominator description. Do not call it the exact number of residents observed for the entire year.
Practice eleven: revised case definition. If surveillance changes from confirmed cases only to confirmed plus probable cases, preserve the break in definition. An apparent increase may not be directly comparable across the change.
Practice twelve: duration. High prevalence can reflect long duration as well as high incidence. Do not translate prevalence as “new cases” simply because the percentage is large.
27. Frequently asked translation questions
Is incidence the same as risk? Incidence proportion can be interpreted as risk over a specified period in appropriate settings, but incidence rate based on person-time is a different quantity. Preserve the source metric.
Is prevalence a rate? Prevalence is usually a proportion, although everyday and historical writing may call it a rate. Translate the conventional term if needed while preserving the mathematical type.
Can person-years be converted into number of people? Not uniquely. Person-years sum observation time. Many different combinations of people and follow-up duration can produce the same total.
Can a rate per 100,000 be written as a percentage? Only if the underlying quantity is a compatible proportion and the time structure is preserved. A person-time rate should retain its person-time unit.
Why can prevalence stay high when incidence falls? Existing cases can remain in the population for a long time. The relationship depends on duration, recovery, mortality, migration and other dynamics. Translate the source explanation rather than applying a simple shortcut automatically.
What is the best release check? For every epidemiological number, write the numerator, denominator and time window in plain language. If the target sentence implies the same three elements, the metric has probably been translated correctly.
28. A release checklist for epidemiological frequency
- Distinguish new cases from all existing cases.
- Preserve point prevalence versus period prevalence.
- Preserve incidence proportion versus person-time incidence rate.
- Keep person-years, person-months or other time-at-risk units attached.
- Preserve case definitions and classification levels.
- Preserve the population at risk or denominator population.
- Keep scaling factors such as per 1,000 or per 100,000.
- Distinguish attack rate from person-time rate.
- Distinguish mortality rate from case-fatality proportion.
- Keep crude, age-specific and standardised rates separate.
- Preserve the named standard population where relevant.
- Keep recurrent-event counts distinct from number of people affected.
- Preserve surveillance or testing changes that affect comparability.
The checklist works because epidemiological translation is denominator translation. Once the case definition, population and time frame are secure, the target language can be concise without becoming vague.
29. Continue through the established eduKate translation architecture
This article sits beneath Master Art of Translation and complements Probability, Odds, Risk Ratios and Absolute Risk, Percentages, Percentage Points and Basis Points, and Sensitivity, Specificity, PPV and NPV. Each owner answers a distinct search intent while sharing a common discipline of preserving denominators and evidence.
The Vocabulary Learning Hub supports distinctions such as new, existing, current, cumulative, crude, adjusted, exposed and at risk. How English Works supports time expressions, quantifiers, prepositions and relative clauses that keep population definitions attached to the right numbers.
The final principle is straightforward: incidence tells a story of new occurrence, prevalence tells a story of existing burden, and person-time tells a story of accumulated observation. Translate the story each denominator encodes. The words can change; the population-time relationship must not.