A patient is being discharged. The nurse explains the new anticoagulant: the dose, what to avoid, which symptoms mean come back immediately. The patient’s teenage son translates. He nods. His mother nods. Everyone signs the form.
Nobody in the room can check what was actually said.
Scenes like this happen every day in hospitals everywhere, and they are usually filed under “communication” or “patient experience.” That is the wrong filing cabinet. The argument here is that language barriers are a patient safety problem, with measurable harm, and that they should be managed with the same seriousness we bring to medication errors and falls.
The harm is measurable
In 2007, Chandrika Divi and colleagues, working with data from hospitals accredited by The Joint Commission, published a pilot study in the International Journal for Quality in Health Care. They examined adverse events in six accredited US hospitals and compared patients with limited English proficiency with English-speaking patients.
The difference in severity was striking. About 49 percent of adverse events among patients with limited English proficiency involved some physical harm, compared with about 30 percent of adverse events among English-speaking patients. When things went wrong for patients who could not communicate easily with their clinicians, they tended to go more seriously wrong.
It was a pilot study, in a limited number of hospitals, and it should not be over-interpreted. But the direction is consistent with what clinicians see: miscommunication about allergies, medication doses, symptoms and consent creates room for exactly the kinds of errors that cause harm.
Professional interpreters work
The encouraging part is that we know what helps. A 2007 systematic review by Leah Karliner and colleagues in Health Services Research examined whether professional interpreters improve clinical care for patients with limited English proficiency. It found that professional interpreters were associated with benefits in communication, including fewer errors and better comprehension, as well as in utilisation, clinical outcomes and satisfaction. The review concluded that care for patients using professional interpreters can approach the quality of care for patients without language barriers.
This is not a fringe intervention. It is one of the clearest-evidence fixes available for a common safety risk.
Why ad hoc interpreting persists
So why are family members, bilingual cleaners and a few words of phrasebook Spanish still doing so much of the work?
The honest answer is that ad hoc interpreting is free and immediate, and professional interpreting costs money and takes time to arrange. In a busy ward, the path of least resistance is whoever is standing nearby.
That shortcut carries specific risks.
- Family members edit. They may soften bad news, omit embarrassing questions or answer on the patient’s behalf. Sometimes that is protective. Clinically, it is distortion.
- Children should not carry adult burdens. Asking a child to interpret a cancer diagnosis or a question about domestic violence is unfair to the child and unsafe for the patient.
- Untrained bilingual staff lack medical vocabulary in both languages and are pulled away from their own jobs.
- Nobody is accountable for accuracy. A professional interpreter works to a code of conduct. A relative does not.
The budget line problem
Here is the uncomfortable incentive. In most hospitals, interpreting services sit in a budget line that looks like overhead. Their costs are visible and immediate. Their benefits, fewer adverse events, shorter stays, fewer readmissions, show up somewhere else, later, and are hard to attribute.
That makes interpreter services an easy target in cost-cutting rounds. A manager who trims the interpreter budget sees the saving this quarter. The patient harmed by a misunderstood discharge instruction appears in a different department’s figures next year, if the link is ever made at all.
This is exactly the pattern patient safety has spent decades trying to break in other areas. We do not treat hand hygiene or surgical checklists as optional extras to be cut when budgets are tight. We should not treat language access that way either.
Technology helps, and has limits
Video and telephone interpreting have made professional interpreters far more available, including for less common languages and at night. That is real progress.
Machine translation is the newer temptation. It is fast, free at the point of use and increasingly fluent. For simple, low-stakes exchanges, it may help. But for consent, diagnosis, medication instructions and end-of-life conversations, an unvalidated tool that can produce confident errors is not a safe substitute. Health services adopting it should do so with clear limits on where it can be used and with evidence from their own setting.
What safe language access looks like
If language barriers are a safety issue, the response should look like other safety programmes.
- Record language needs at registration, and flag them in the record like an allergy.
- Make professional interpreting the default for clinical decisions, consent and discharge, with ad hoc interpreting reserved for genuine emergencies.
- Use teach-back through the interpreter to confirm understanding, rather than accepting a nod.
- Report language-related incidents in safety reporting systems, so the problem becomes visible in the data.
- Fund interpreting as clinical infrastructure, not as a discretionary add-on.
Where the error enters
It helps to be specific about where language barriers produce harm, because the risks cluster at predictable moments.
Allergy and medication history. At admission, a patient is asked about allergies and current medicines. Without a competent interpreter, “I get a rash from that antibiotic” can become “no allergies.” A herbal remedy or a medicine bought abroad may go unmentioned. The error is silent until a reaction occurs.
Symptom description. Pain, breathlessness, dizziness and confusion are described differently across languages and cultures. A family member may translate “pressure in my chest” as “a bit of discomfort.” Triage decisions follow the translation, not the symptom.
Consent. A consent conversation that passes through an untrained interpreter may lose the risks, the alternatives or the option to refuse. The signature on the form records agreement, not understanding.
Discharge. This is often the most dangerous point. New medicines, changed doses, warning signs and follow-up arrangements are all delivered at once, frequently in a hurry, to a patient who is tired and eager to go home.
Across all four, the common feature is that errors are invisible in the moment. Everyone believes communication has taken place. Only the outcome reveals that it has not.
The strongest counterargument
The most serious objection to making professional interpreting the default is practical rather than principled.
Professional interpreters are not always available, especially for less common languages, in rural areas or at night. Waiting for one can delay care, and in urgent situations delay can be dangerous. Family members bring real benefits: they know the patient’s history, they provide emotional support, and patients often prefer them. Some patients actively refuse a stranger in the room when discussing intimate matters. And budgets are finite. Every pound or dollar spent on interpreting is not spent on something else.
Much of this holds. In genuine emergencies, clinicians must use whatever communication is available, and nobody should delay life-saving treatment to wait for an interpreter. Family members have an important role in supporting patients, and patient preferences matter.
But the counterargument supports a narrower conclusion than it seems to. It supports flexibility in emergencies and respect for patient preferences. It does not support using untrained interpreters routinely for planned conversations, such as consent for elective surgery or discharge planning, where there is time to arrange a professional. Telephone and video interpreting have also removed much of the availability problem for common languages. And where patients prefer a family member, the safest approach is often to have both: the family member for support, the professional for accuracy.
On cost, the relevant comparison is not interpreting versus nothing. It is interpreting versus the cost of the errors, longer stays and readmissions that follow miscommunication, which are real even if they sit in different budget lines.
Second-order effects
Clinicians practise defensively. When clinicians cannot take a reliable history, they may order more tests or admit patients they would otherwise send home, simply to compensate for uncertainty. That increases costs and exposes patients to unnecessary procedures.
Children lose. Children who interpret for their parents miss school, carry emotional burdens and are exposed to information no child should have to relay. That is a harm in its own right.
Trust erodes. Patients who repeatedly leave appointments unsure of what was said learn that the health system is not built for them, and may avoid it until problems are serious.
Research excludes. Many clinical trials exclude participants who do not speak the dominant language, which means evidence is weakest for the populations that already face the greatest barriers.
What this means for different decision-makers
For health ministries and regulators, language access can be built into accreditation and quality standards, with clear expectations about when professional interpreting is required. Commissioning bodies can fund interpreting as part of the cost of care rather than as a separate, easily cut service.
For hospital leaders, the most useful steps are operational: making telephone and video interpreting easy to access on every ward, tracking how often professional interpreters are used for consent and discharge, and treating failures as reportable safety events.
For educators, clinical training should include how to work with interpreters: speaking directly to the patient, using short segments, avoiding jargon and checking understanding. These are skills, and they are rarely taught.
For clinicians, the habit worth building is to ask, early and explicitly, what language the patient prefers for medical conversations, and to plan for it, just as one would plan for a known allergy.
What to watch next
Watch how machine translation and AI-based tools are introduced into clinical settings. They may become useful for low-risk tasks, but evaluation in real clinical encounters, across many languages, will matter far more than vendor claims. Watch whether patient safety reporting systems begin to capture language as a contributing factor, which would make the scale of the problem visible. And watch for research that includes patients with limited proficiency in the dominant language, rather than excluding them by default.
Sign language and other overlooked languages
Discussions of language access usually focus on spoken languages from migration. Two groups are often overlooked.
Deaf patients who use sign language face many of the same risks, and sometimes worse ones, because staff may assume that written notes are an adequate substitute. For many Deaf people, the national sign language is their first language and written text in the spoken language is a second language. Notes passed across a bed are not a safe way to obtain consent or explain a new diagnosis.
Speakers of Indigenous and minority languages within a country can also be missed, because services assume that everyone born there speaks the dominant language fluently. Older patients in particular may be far more comfortable in their first language, especially when they are unwell, frightened or cognitively impaired.
A language access policy that covers only the most common migrant languages is incomplete. The safety logic applies to everyone who cannot communicate reliably in the language of the consultation.
The takeaway
When a patient cannot understand their clinician, every other safety system is weakened. Checklists depend on accurate histories. Medication reconciliation depends on the patient understanding what they take. Consent depends on comprehension.
Treating language access as a courtesy rather than a safety requirement means accepting a higher rate of serious harm for a large group of patients. The evidence on professional interpreters is clear enough that this is now a choice, not an unknown.
How does your service handle interpreting for discharge and consent conversations? Tell us what works.
