How do you translate a public health surveillance report, outbreak case definition or epidemiological dashboard accurately? Preserve the classification rules, denominators, dates and uncertainty before improving the sentences. Public health translation is not only a matter of finding equivalents for medical terms: it must protect what a statistic counts, who is eligible to be counted, which observation period applies, and whether a case is suspected, probable, laboratory-confirmed or excluded under the particular surveillance definition.
This guide explains the top ways to translate epidemiology data, infectious-disease surveillance tables, outbreak investigation line lists, case definition criteria and public health dashboard labels without changing their evidence. It is designed for translators, public health students, educators, data teams and multilingual communicators. You will find first principles, a repeatable method, worked examples, failure modes, a practice lesson and review questions for making technical material readable without misrepresenting the underlying evidence.
The difference between a sound and an unsafe translation may be only one word: ‘reported’ instead of ‘onset’, ‘suspected’ instead of ‘confirmed’, ‘per 100,000 residents’ instead of ‘per resident’, or ‘less than five’ instead of ‘five’. The numbers themselves can remain identical while the interpretation becomes entirely wrong. That is why this article treats epidemiological translation as a problem of evidence preservation and not a competition for the smoothest paragraph.
The visible problem: an accurate-looking chart can tell the wrong story
Imagine a multilingual dashboard that displays two dates for each event: symptom onset and report receipt. A translator chooses ‘date of infection’ for both. A reader then interprets the rise in reported cases as proof of a rise in infections on precisely those dates. In fact, the underlying records may reflect reporting delays, retrospective entry or changes in testing access. The translated heading created a claim that was never contained in the data.
A second error is more subtle. A report uses ‘suspected cases’ for an operational category under investigation. A translation calls them ‘confirmed infections’. Even when the count remains 82, the translated headline implies a different level of evidence. People reading translated public health information may make decisions from that headline. Misclassification therefore matters more than whether the translated wording resembles everyday medical conversation.
Finally, a dashboard may show a rate per 100,000 population and a separate absolute count. Translation can accidentally collapse them into one ‘number of cases’ measure. Rates support comparisons under particular denominator choices; counts answer how many events were recorded. Confusing the two can reverse a reader’s conclusion about which area experienced a greater burden.
First principles: understand what surveillance measures
A case definition is a rule for classifying events for a specified purpose, disease and setting. It may include clinical findings, epidemiological links, laboratory evidence, timing, geography and exclusions. The labels ‘suspected’, ‘probable’ and ‘confirmed’ are not universal clinical stages with identical criteria across diseases or countries. They are defined categories that must be translated from the issued document, with its date and version. A translator should not substitute a familiar definition from another outbreak.
A line list is a structured row-level operational record. Each row may represent a person, episode, investigation, specimen or event depending on the programme. One person can have more than one test and some records may be duplicates pending reconciliation. A translated heading such as ‘patients’ is wrong if the table actually counts specimens. The unit of observation is as important as the field name.
Incidence, prevalence, positivity, mortality and case fatality have different numerator and denominator structures. The translator’s job is to preserve the measure actually calculated and its limitations. A case fatality proportion among reported cases is not automatically the chance of death after infection; a test positivity fraction is not the fraction of the whole population infected. These distinctions must remain visible even when the public-facing summary uses plain language.
Time in surveillance has several meanings: symptom onset date, specimen collection date, test result date, notification date, report date and publication date may all differ. Epidemiological weeks may follow a specific national or international convention. Never replace these with a generic ‘infection date’ because it sounds intuitive. A comparison across countries may require explaining that definitions and reporting arrangements differ.
Uncertainty is a necessary part of responsible communication. Preliminary counts may change after deduplication or laboratory reconciliation; model estimates are not direct observations; confidence intervals describe uncertainty under a statistical method; suppressed values protect confidentiality and may be reported as intervals. A translation that removes these statuses can produce more apparent certainty than the source warrants.
A reliable ten-step workflow for multilingual epidemiology
- Identify the source authority, surveillance objective, publication date, case-definition version, geographic coverage and reporting period.
- List each unit of observation: person, case, episode, visit, specimen, positive test, contact, death or notification.
- Record the exact eligibility and classification rules, including and/or conditions, exclusions, time windows and laboratory criteria.
- Build a field dictionary with original label, approved translation, definition, data type, allowed values, unit and example.
- Freeze codes, dates, counts, denominators and suppression conventions before translating explanatory prose.
- Separate event date from reporting date and describe the exact time convention used by each chart.
- Translate suspected, probable, confirmed and excluded only with the definitions governing this dataset.
- Check numerator, denominator, unit and population scope for every rate, percentage, comparison and caption.
- Perform privacy review: do not expose personal identifiers or reconstruct confidential small-cell information through the translation.
- Arrange independent language, epidemiology, statistical and publication reviews with a traceable decision log.
The workflow deliberately starts with the source’s purpose. A school health survey, a laboratory surveillance bulletin and an active outbreak line list do not necessarily define their rows or categories in the same way. The translator who reads the dataset’s ‘about’ page before the headlines is less likely to use the wrong everyday synonym. A change that affects classification, statistics or public action must be resolved by the relevant domain owner, not by silent editorial judgement.
Use a two-column interpretation ledger as a quick habit: on the left, record exactly what the source claims; on the right, record what a reasonable reader could infer from the proposed translation. If those meanings differ, revise the target text. This step catches dangerously fluent errors that a spelling or grammar checker cannot detect.
Twenty-four worked cases: diagnose the precise translation failure
The cases below are fictional teaching exercises. They do not describe actual outbreaks, people or current disease totals. The short source fragments are English glosses of hypothetical records, not quoted public health bulletins. Genuine translation work must use the controlling disease- and location-specific definitions, verified data tables and approved operational procedures.
Suspected is a defined category
Teaching source: Dashboard: 82 suspected cases under the current surveillance definition. Faulty translation: The headline reads ’82 confirmed infections’. Evidence-preserving translation: There are 82 reports classified as suspected cases under the stated definition.
The classification shows what evidence is available at a particular stage. It does not guarantee later confirmation or mean that the people have no symptoms. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Check the document’s criteria and keep the exact category label. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Explain why a category name cannot be replaced solely to make a headline more dramatic.
Probable does not mean mathematically likely
Teaching source: Case category: probable, following specified laboratory and epidemiological criteria. Faulty translation: The text translates probable as ‘more than 50% likely’. Evidence-preserving translation: Classified as a probable case according to the issued surveillance definition.
An operational category is not automatically a probability. Converting it into a numerical chance invents a meaning the definition may not contain. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Quote the case-definition version in the translator note. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Write a plain-language explanation that does not add an estimated probability.
Confirmed depends on a rule
Teaching source: Case classified as confirmed following the specified confirmatory process. Faulty translation: The translation says ‘patient has fully recovered’. Evidence-preserving translation: Case classified as confirmed under the current criteria.
Confirmation refers to classification evidence, not recovery or clinical severity. A category can be accurate without implying anything about the person’s later outcome. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Separate diagnosis or surveillance classification from outcome fields. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: List which later statuses would require separate source evidence.
Possible case does not always exist
Teaching source: Local reporting scheme includes suspected, probable and confirmed only. Faulty translation: An extra possible-case category is created for unfamiliar records. Evidence-preserving translation: The reporting categories are those specified by the local surveillance scheme.
Some words sound like natural synonyms but are not registered categories. Adding an unapproved value can break data matching. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Compare allowable code values before translation. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Create a field dictionary with no invented classifications.
Inclusive and exclusive logic
Teaching source: Eligible if criterion A or B and exposure condition C is satisfied. Faulty translation: The translated rule requires A and B and C. Evidence-preserving translation: Eligible where the stated combination of A or B, together with C, is met.
Logical connectors govern which events qualify. Changing ‘or’ to ‘and’ changes the counted population, even if each criterion is translated correctly. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Rewrite eligibility as a parenthesised logical test. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Construct two hypothetical examples that expose the difference.
Symptom onset is not report receipt
Teaching source: Onset 4 June; notification 9 June. Faulty translation: Both fields become ‘infection date’. Evidence-preserving translation: Symptom onset recorded on 4 June; notification received on 9 June.
The source names distinct observed or administrative milestones. Neither date alone proves when infection occurred. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Retain date-field labels through tables and charts. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Show how a reporting delay can affect an epidemic curve.
Specimen date versus result date
Teaching source: Specimen collected 12 July; result issued 14 July. Faulty translation: The later date becomes collection date. Evidence-preserving translation: Specimen collection occurred on 12 July; laboratory result was reported on 14 July.
The dates answer different questions and may follow separate workflows. Collapsing them can create false timeliness metrics. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Cross-check field definitions before publishing. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Write a timeline with all events labelled rather than merged.
Reporting week convention
Teaching source: Week 01 under a declared epidemiological-week calendar. Faulty translation: Week 01 is assumed to mean January 1–7. Evidence-preserving translation: Epidemiological week 01 under the stated reporting calendar.
Week boundaries can differ from ordinary calendar weeks and conventions may differ. A translation must preserve the original time grouping. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Include the definition or reference for the week calendar. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Explain why weekly totals cannot always be compared without a shared convention.
New and cumulative counts
Teaching source: This week 15 new reports; cumulative total 210. Faulty translation: The heading calls 210 new reports. Evidence-preserving translation: Fifteen newly reported events this week; cumulative total 210 over the stated period.
A cumulative figure aggregates earlier periods. It cannot be compared as though it were a single-week count. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Match every figure to the period named in the source. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Calculate which number belongs in a week-over-week comparison.
Cases versus tests
Teaching source: 340 tests processed; 28 returned positive results. Faulty translation: The bulletin reports 340 infected people. Evidence-preserving translation: Three hundred forty tests were processed, of which 28 produced positive results.
Tests, positive tests and unique people are not the same count. Individuals may contribute multiple tests, and further classification may be required. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Confirm the table’s unit of observation. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Rewrite the headline without implying a unique-person count.
Deduplication status
Teaching source: Daily numbers preliminary, before duplicate-record reconciliation. Faulty translation: The dashboard claims final certified totals. Evidence-preserving translation: Daily counts are preliminary and may change after duplicate records are reconciled.
Data-cleaning status matters whenever the same event can appear through multiple reporting pathways. A translated footnote should keep that possibility visible. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Track provisional and final releases separately. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Describe what can change without claiming the underlying outbreak itself changed.
Incidence versus prevalence
Teaching source: New cases during the defined period per population at risk. Faulty translation: The figure is called the share of all people currently ill. Evidence-preserving translation: Incidence measure for new events during the specified time interval.
Incidence concerns new occurrence over time; prevalence concerns existing cases at a point or period. Their denominators and interpretations differ. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Write the numerator, denominator and time interval beside the term. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Explain why the two measures answer different public health questions.
Rate denominator
Teaching source: Rate 24 per 100,000 residents over the reporting period. Faulty translation: The translation reports 24% of residents. Evidence-preserving translation: Reported rate: 24 events per 100,000 residents for the stated period.
A per-100,000 rate cannot be replaced with a percentage by preserving only the digits. A 24% claim would be fundamentally different. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Recalculate only to check the interpretation, not to replace the official published figure. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: State the unit and population in one complete sentence.
Age-standardised versus crude rate
Teaching source: Age-standardised rate displayed beside crude rate. Faulty translation: Both are labelled ‘average rate’. Evidence-preserving translation: Age-standardised rate and crude rate, each as separately reported.
Standardisation changes how population structure enters a comparison. Translators should not remove its label merely because it is technical. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Keep the method explanation or link supplied by the issuer. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Ask what comparison the reader is entitled to make with each measure.
Positivity is not population prevalence
Teaching source: Twenty-eight positive results among 340 processed tests. Faulty translation: The dashboard says 8.2% of residents were infected. Evidence-preserving translation: Approximately 8.2% of processed tests were positive in the stated testing set.
The denominator is the selected tests, not the population. Who sought or received testing affects the interpretation. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Preserve the tested population and period. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Rewrite a misleading media summary without adding an unmeasured population claim.
Confidence interval
Teaching source: Estimated rate 12.4; 95% interval 9.1–16.3, as defined in source. Faulty translation: The interval is described as a guaranteed range for future counts. Evidence-preserving translation: Estimated rate 12.4 with a reported 95% confidence interval of 9.1–16.3.
An interval is tied to a statistical construction and its assumptions. It does not promise that future observations must fall inside its numeric bounds. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Retain the stated confidence level and method label. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Write a cautious interpretation without claiming certainty.
Observed versus modelled values
Teaching source: Projected range based on modelling, not a count of reported cases. Faulty translation: The projection becomes a tally of confirmed cases. Evidence-preserving translation: Model-based projection for the stated scenario and period.
Forecasts are conditional estimates that can change with assumptions, while observed surveillance counts arise from reporting systems. These must not be merged. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Mark charts as observed, estimated or projected. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Show why a model line should not inherit the count label of the adjacent table.
Small-cell suppression
Teaching source: A geographic cell is suppressed because the count is below the publication threshold. Faulty translation: The number is replaced with an invented estimate. Evidence-preserving translation: Count suppressed under the reporting authority’s disclosure rule.
Suppression protects privacy and sometimes data reliability. A translation should preserve the symbol and official explanation, not reconstruct hidden values. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Check the suppression legend and any cumulative disclosure risk. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Write a caption that explains omission without exposing small-cell identities.
Named location and privacy
Teaching source: District-level count published; residential addresses not released. Faulty translation: The translation adds a street-level map from inferred addresses. Evidence-preserving translation: Events reported at district level as released by the authority.
Public communications should not enrich a dataset with sensitive geographic detail that its source intentionally excludes. The correct granularity is part of responsible translation. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Apply the same disclosure boundary to translated visualisations. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Explain why a more specific map is not automatically a more faithful translation.
Contact versus case
Teaching source: Contact list includes persons under follow-up who are not classified as cases. Faulty translation: All contacts become infected cases. Evidence-preserving translation: Contacts under follow-up, recorded separately from classified cases.
Exposure or monitoring status does not establish infection. The distinction matters for both risk interpretation and numbers. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Check allowed values for follow-up and case-status columns. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Draft a two-sentence clarification distinguishing these populations.
Imported versus locally acquired
Teaching source: Case investigation records travel-associated acquisition as suspected. Faulty translation: The translation states a confirmed local transmission chain. Evidence-preserving translation: Investigators suspect travel-associated exposure; acquisition route remains under review.
Source attribution can be preliminary, mixed or unresolved. Never turn an investigation hypothesis into a geographic certainty. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Preserve evidence status and investigation date. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: List what additional evidence would support a stronger claim.
Deaths with and deaths from
Teaching source: Report uses a defined category ‘deaths among reported cases’. Faulty translation: The target says every death was caused by the infection. Evidence-preserving translation: Deaths among reported cases, according to the publication’s definition.
Association, attribution and causal certification may be separate processes. A translator must preserve the exact classification used by the source. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Consult the report’s methodology and certifying definitions. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Rewrite a headline to prevent ungrounded causal inference.
Retrospective revision
Teaching source: Previous week’s figure revised from 63 to 59 after record review. Faulty translation: The revision is framed as a decrease of four new infections. Evidence-preserving translation: The previously published figure was revised from 63 to 59 after record review.
Revision to a historical tally is not necessarily a new epidemiological event. The reason and version date matter. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Link each number to its release and revision. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Explain how a revised series differs from a live daily change.
Sampling caveat and representativeness
Teaching source: Sentinel sample includes selected clinics, not all residents. Faulty translation: The translation claims a complete population census. Evidence-preserving translation: Results are from the specified sentinel clinics and should be interpreted within that coverage.
Selection affects what the data can represent. A translation should retain eligibility and coverage limitations even when a short caption is preferred. The examples make clear that trustworthy translation depends on understanding how the data were produced and classified, not merely learning a bilingual list of terminology. Write the target wording for its intended reader, but keep the source’s scope, qualifications and relationships intact. An explanatory parenthesis is acceptable when it clarifies an existing rule; introducing a new measurement or diagnosis is not.
Copy scope restrictions into the translated summary. If a term remains ambiguous, record an explicit query for the relevant public health or data owner. Such questions should identify the original table or field, the candidate meaning, the proposed target phrase and why it matters to a reader. Practice: Explain why an unqualified national claim would exceed the evidence.
How to translate AND, OR, NOT and conditional criteria
Case definitions commonly combine different criteria. Language makes these combinations easy to misread because natural-language ‘and’ or ‘or’ can have scope that is not obvious from a long sentence. Before translating the paragraph, create a small logical outline: identify mandatory conditions, alternative pathways, exceptions and exclusions. Do not change the logic to make the sentence shorter.
For example, a hypothetical classification may require either a specified clinical presentation or a laboratory marker, plus a stated exposure window. The reader needs to know which requirements apply to both branches and which apply only to one branch. A table or diagram can supplement an awkward sentence, but should not replace the authorised original. Ask a subject reviewer to test boundary cases: a record satisfying only the first condition, only the second, both, or neither.
Negation demands special attention. ‘No evidence of exposure’ does not always mean ‘evidence of no exposure’. ‘Test not performed’ differs from ‘test negative’; ‘result unavailable’ differs from ‘result pending’. Each phrase corresponds to a different state in a data system. If the target language uses a different grammatical structure, preserve the distinction explicitly rather than leaning on a single vague word.
Keeping people safe while making technical text understandable
The most useful public health translation is accessible without exaggeration. Define a specialist term once, keep the official category in view and explain what readers should and should not infer from the figure. Do not create behavioural advice, medical recommendations, isolation instructions or emergency thresholds that are absent from the issuing authority’s source. If the source contains current instructions, link the reader to the authority’s most recent guidance rather than republishing a potentially stale rule as universal advice.
Line lists can contain sensitive personal information. Translators and editors should work with access controls, confidentiality commitments and the minimum data needed for their task. Training exercises can use invented data. Public dashboards may deliberately suppress granular counts or locations; translations should not circumvent those choices by combining fields to infer identities. The capacity to reconstruct private information is not a sign of translation accuracy.
A multilingual release also needs version control. State the data extraction or publication date, jurisdiction, language version and translation contact. When a case definition is revised, treat it as a substantive source change: assess the translated fields, examples and dashboard labels again. A language change that appears minor may invalidate comparisons that previously depended on an old classification rule.
Practice: teach the method with a fictional surveillance bulletin
Give learners a one-page fictitious bulletin containing a case definition, 12 rows of invented line-list data, an onset-date curve, a report-date curve and rates for two districts. Ask students first to annotate every denominator, classification, evidence status and date field. Then ask them to produce a short target-language summary for non-specialist readers and a precise field dictionary for technical users.
The reviewer should challenge the translation with a few adversarial questions. Does the sentence treat tests as people? Does a suspected category become confirmed? Does the weekly chart use onset or report date? Is a count being compared with a rate? Did a suppressed cell become a guessed value? Ask the learner to point to the original field supporting each answer. The stronger the method, the less the reviewer needs to rely on the translator’s confidence.
A second round asks the student to transfer the principle beyond health surveillance. A school assessment report also has defined cohorts and denominators; an oceanographic dataset has observed and processed quantities; a financial report distinguishes forecasts and actuals. In every case, accurate translation requires a reader to recover the same underlying measurement contract.
Frequently asked questions
Are suspected, probable and confirmed universal definitions?
No. Their criteria are specified for each surveillance system and disease, and may vary by time and purpose. Consult the exact source definition rather than assuming a global set of interchangeable labels. Where the reporting situation is current or high stakes, refer readers to the responsible public health authority for the controlling version.
Does a positive test always equal a counted confirmed case?
Not necessarily. Case classification may depend on the applicable rules, the test type, repeat testing, episode definitions and reconciliation procedures. Translate the field as the data source defines it. Where the reporting situation is current or high stakes, refer readers to the responsible public health authority for the controlling version.
What is the difference between incidence and prevalence?
Incidence addresses new occurrence over a defined time and population, while prevalence describes existing cases at a point or during a period. They have different interpretations and must not share a vague label when precision matters. Where the reporting situation is current or high stakes, refer readers to the responsible public health authority for the controlling version.
Why are onset and notification dates different?
Symptoms can begin before a report is filed or received. Reporting delays and revisions can change a notification-based chart without changing the original symptom-onset events. Where the reporting situation is current or high stakes, refer readers to the responsible public health authority for the controlling version.
Can I translate a rate as a percentage?
Only if the conversion is mathematically valid, explicitly authorised and does not hide the source denominator. A rate per 100,000 is not the same thing as a percentage displayed with identical digits. Where the reporting situation is current or high stakes, refer readers to the responsible public health authority for the controlling version.
What does a suppressed count mean?
It means the publisher has withheld a value under a specified disclosure or reporting rule. Preserve the suppression and its legend; never infer a hidden number solely to produce a fuller-looking translation. Where the reporting situation is current or high stakes, refer readers to the responsible public health authority for the controlling version.
Can translation software check an epidemiology dashboard?
It may help identify inconsistent terms and missing fields, but it cannot certify case definitions, denominators or confidentiality safeguards. Domain review and comparison with the source data remain necessary. Where the reporting situation is current or high stakes, refer readers to the responsible public health authority for the controlling version.
How do I handle a contradictory source table and summary?
Document the mismatch with its locations, versions and possible interpretations. Ask the issuing authority or authorised data owner to resolve it rather than silently editing the numbers or choosing the more attractive claim. Where the reporting situation is current or high stakes, refer readers to the responsible public health authority for the controlling version.
Reference material and limits
The World Health Organization’s case-definition reference demonstrates how suspected, probable and confirmed categories depend on stated clinical, epidemiological and laboratory conditions. The WHO measles outbreak toolbox illustrates the role of disease-specific definitions. These sources provide conceptual grounding, not medical instructions for the invented examples in this guide.
Where to learn next
For general translation judgement, read Translate Meaning, Not Just Words and Context, Tone and Register. For controlling official definitions and labels, use Terminology, Glossaries and Quality Checks.
The Top Ways to Translate Correctly series index connects other specialist contexts. The Vocabulary Learning Hub, How English Works: Translation and Meaning Across Languages and Sources, Attribution and Verification help students understand the language and evidence skills behind dependable explanations.
The principle to remember
Translate not just the label but the rule under which the label is true. Epidemiological communication becomes trustworthy when readers can identify what was counted, who was counted, when, where, by which method and with what uncertainty. The words should be accessible; the statistics must remain honest. That is the foundation of accurate translation across public health languages.