Epic Now Delivers Discharge Docs in Your Language — But Which?

A Real Step Forward — With a Significant Blind Spot

Language Services Associates just announced HealthTranslate, an integration that pushes translated discharge instructions and after-visit summaries directly through Epic's MyChart. For limited English proficient (LEP) patients, that is genuinely useful. No more waiting for a staff member to track down a translated handout. No more sending patients home with instructions they cannot read.

But here is the question nobody in the press release asks: which languages are actually covered?

For Spanish, Mandarin, Arabic, and the other top-ten LEP languages in the US, platforms like this work reasonably well. For Chuukese and Pohnpeian speakers — two of the fastest-growing Pacific Islander communities in states like Hawaii, Guam, Arkansas, and Oregon — integrations like HealthTranslate are almost certainly silent. That silence has clinical consequences.

Why EHR Language Integrations Skip Micronesian Languages

The business logic is straightforward. Vendors build for volume. If a health system has 40,000 Spanish-speaking patients and 300 Chuukese-speaking patients, the API gets trained, tested, and quality-checked in Spanish. Chuukese gets a workaround, if it gets anything at all.

The problem is that the 300 Chuukese patients are not evenly distributed. In certain zip codes — parts of Honolulu, Springfield (Missouri), and Portland — Chuukese speakers can represent 15 to 20 percent of a clinic's patient population. For those clinics, the HealthTranslate integration solves the language access problem for everyone except the patients who already had the fewest resources.

Chuukese and Pohnpeian are Austronesian languages with no meaningful machine translation support as of 2025. There is no production-ready neural MT engine for either language. That means any Epic-integrated workflow that relies on AI or statistical MT to generate translated discharge summaries will either skip these languages entirely or produce output that is dangerous to use clinically.

What Clinical Accuracy Actually Requires for These Languages

Discharge instructions are not marketing copy. A mistranslated phrase in a brochure is embarrassing. A mistranslated dosing instruction or follow-up appointment window can cause a readmission — or worse.

For Chuukese and Pohnpeian, clinical translation quality control requires a few things that generalist EHR integrations rarely build in:

Terminology consistency. Neither language has a standardized medical glossary. A Chuukese translator working on cardiology materials needs to use the same term for "blood pressure" that the patient's interpreter used during the appointment. Without a project-specific glossary locked to the health system's preferred terminology, you get inconsistency across documents — confusing for any patient, and dangerous for someone managing a chronic condition.

Back-translation or bilingual review. Because there are no MT post-editing benchmarks for Chuukese or Pohnpeian, human back-translation is the most defensible quality gate. A second qualified linguist re-translates the output into English without seeing the source, and a clinical reviewer compares the two. This adds time and cost, but for discharge instructions it is non-negotiable.

Cultural routing, not just linguistic routing. Pohnpeian and Chuukese communities have distinct relationships with authority, family decision-making, and disclosure of diagnosis. A discharge summary addressed to an individual patient may land differently than the same information framed for a family unit. Translation alone does not account for that — interpreter briefing and document design do.

How the Gap Shows Up in Practice

What Health Systems Should Ask Their Language Vendors Right Now

If your health system is adopting HealthTranslate or any Epic-integrated translation tool, three questions will tell you whether you have a real solution or a partial one:

1. Which languages are covered by machine translation, and which require human-only workflows? Get the list in writing. If your vendor cannot name the specific languages that go through human review rather than MT, that is a red flag.

2. How does the integration handle a language it cannot process? A good integration fails visibly — it flags the document as untranslated and routes it to a human workflow. A bad one silently returns nothing, and the patient goes home without instructions.

3. Who are your credentialed Chuukese and Pohnpeian translators? This is not a trick question. If your community includes Pacific Islander patients under the Compact of Free Association (COFA), those patients have federally documented rights to language access. If your vendor cannot name qualified linguists for those languages, they cannot serve that population.

The Bigger Picture

LSA's HealthTranslate integration is a meaningful product for the majority of LEP patients in most US health systems. Streamlining multilingual discharge documents through existing clinical workflows removes real friction and supports compliance with Section 1557 of the Affordable Care Act.

But language access policy measures averages, not outliers. The communities with the worst health outcomes — Pacific Islander populations, recent COFA migrants, rural Micronesian communities — are exactly the ones least likely to be covered by a scalable API integration built for high-volume languages.

If you serve those communities, the integration is not enough on its own. You need a subcontracting relationship with a vendor who has actual Chuukese and Pohnpeian capacity — and who can plug into your Epic workflow when the platform's built-in coverage stops.

That is a specific vendor requirement, not a philosophical position. Ask your current LSP to prove it before the next discharge summary needs translating.

If you want to talk through how rare-language translation fits into an Epic environment, reach out to TXLOC — we work directly with health systems and LSPs on exactly this.