Assuring the quality of a medical translation
Medical translation quality assurance combines qualified linguists, bilingual revision, final checks and continuing oversight of the process. Human and automated checks complement each other, while commissioning organisations need reviewers who understand both the language and the clinical or regulatory content.
Translator qualifications
Quality assurance aims to deliver a translation that meets the requester's requirements. Incorrect dosage instructions can cause medication errors, and poorly translated device instructions can cause incorrect use, with patient harm, costs and reputational damage for providers and clients. Both human and automated quality assurance have a place, but fully automated checks cannot suffice where the source meaning must be rendered accurately.
Suitably experienced linguists are essential, although a medical qualification is not a prerequisite. For jargon-heavy discharge letters, medical reports or descriptions of surgical procedures, a medically trained translator can help; otherwise a third review by a subject-matter expert may suffice. A technical or engineering background can help with medical-device texts.
Bilingual revision and final checks
Bilingual revision compares source and target side by side, regardless of whether a CAT tool is used. Translators should first check their own work and, wherever possible, have another medical translator review it bilingually. ISO 17100 includes bilingual revision as a workflow requirement. A language service provider organises revision for its contractors; translators working directly with patients or healthcare professionals should consider arranging it with colleagues.
A completed translation should receive a word-for-word bilingual check by a second translator at least as qualified or experienced as the first. Some providers use junior translators followed by senior medical revisers, and returning the final quality-controlled document to the junior translator can support learning. Where few suitably expert native speakers are available, as for Baltic languages or Hungarian into English, non-native translators and revisers may do the work; a subsequent review by a native English speaker with subject expertise is strongly recommended to check terminology and polish the English. Rushing revision can impair the final translation's quality.
Source ambiguities should be resolved as far as possible, particularly where patient safety is at stake, and remaining doubts highlighted in the translation, with discussion among linguists or with the client where needed. Final checks range from spelling and source-consistent formatting to detailed source–target comparison, depending on the document and provider's process. Numbers in dosages and administration frequency warrant particular attention, alongside mandatory terminology.
Common problems to check during human quality assurance include:
- Prefixed words with opposite meanings — hypertension and hypotension, hyperglycaemia and hypoglycaemia.
- Alternating references in one text — renal and hepatic beside kidneys and liver — getting mixed up in the translation.
- The usual pairing typed in place of the one the source has — in liver or renal impairment, where the source said liver or respiratory tract impairment.
- Copy-and-paste errors, or an accepted memory suggestion, between similar phrases — critically ill adults with fever reused where the source said critically ill children with fever.
- Similar-sounding words — intravascular for intramuscular, with serious safety implications, or pivotal study for pilot study.
- A not dropped, or added, in instructions on how to administer a product or in whom it may be used.
- An item missing from a long list such as adverse reactions, which is why counting through lists helps.
- A seemingly insignificant word lost in a long sentence — for example, “at least” in “… is indicated for the treatment of patients who have progressed after at least two courses of chemotherapy for advanced disease”.
Continuing process controls
In one medical translation agency's 2024 process, translators were tested against hiring criteria before assignment. Internal quality checkers then checked translation-memory and terminology consistency after translation, proofreading and any back translation or master review, both in the CAT tool and in the laid-out document. The agency was testing AI quality-check tools and could not yet report their efficiency.
In 2024 Medtronic's linguistic qualification process required a medical background and native-speaker status. Continuing translation-memory maintenance and term bases also supported linguistic quality. Language leads monitored process adherence, answered linguists' questions and handled project-manager escalations. Linguists received XTM glossaries and language-specific PDF style guides covering punctuation, translation style and segment rejection.
Projects used two linguists unless almost entirely made up of exact matches. Revisers received automated XTM quality reports flagging potentially problematic segments, while a weekly system-wide scan checked ongoing projects for data issues and source mismatches, such as text copied into the wrong source context. Every six months, language leads monitored each freelancer's process adherence, including how many modules changed during revision and whether context files were opened.
Training for company systems
A proprietary content-management system can require knowledge that intermittent freelance users struggle to retain between projects. In 2024 Medtronic traced several major project problems in its internally developed system to inadequate linguist training and inaccessible reference material. An approximately 80-page guide could be skipped under quick-turnaround pressure. The company was developing a focused reference of fewer than ten pages and training using test materials and projects, rather than only documents to read.
Commissioning-side review
Careful vendor evaluation and continuing review are necessary for reliable outsourced translation. In China in 2013, vendor quality varied greatly and some drafts required almost complete in-house reworking, partly because some vendors used part-time student translators.
In Japan in 2013, medical writers were expected to translate, fact-check and ensure consistency between Japanese and English documents. Vendor translators' language expertise did not necessarily include clinical studies, regulatory requirements or expected wording, so writers supplied regulatory input and rewrote translations as needed; finalisation could be time-consuming. Native-speaking regulatory writers' knowledge of clinical trials, required terminology and document content made them suited to checking agency translations in China, with similar language advantages applying in Japan.