Using AI in Healthcare Translation: Best Practices for Hospitals and Providers
Whether it’s being used to detect early signs of disease or bridge language gaps between patients and their doctors, artificial intelligence (AI) is shaking things up in the healthcare industry.
According to recent research from Deloitte, 75% of leading healthcare companies are “experimenting with or planning to scale generative AI across the enterprise,” and even more (92%) see its potential to improve efficiency in healthcare. But although AI can certainly enhance human-led procedures, it’s far from a replacement for human healthcare professionals.
One key strength of AI is its ability to enhance language access measures by speeding up the translation process, helping patients with limited English proficiency (LEP) communicate with their doctors. AI in translation is a powerful tool, but as we’ve seen with machine translation, generative AI-powered translation requires a human in the loop who can ensure linguistic accuracy, cultural appropriateness, and legal compliance.
In this blog post, we’ll take a look at the best practices that hospitals and other institutions should follow when implementing AI in healthcare translation. We’ll also go over some real-world examples of how providers are implementing AI in translation while also maintaining human oversight to ensure accurate and culturally sensitive language access measures.
Table of Contents
Understanding Generative AI in Healthcare Translation
How Are Hospitals & Healthcare Providers Using AI?
Benefits of AI in Healthcare Applications
Risks of AI in the Healthcare Industry
Best Practices for Implementing AI in Healthcare Translation
Future Outlook of AI in Translation and Healthcare
Frequently Asked Questions About Generative AI in the Healthcare Industry
Understanding Generative AI in Healthcare Translation
Generative AI is a form of AI that can create text, images, and videos based on prompts written in natural language. Tools like ChatGPT and Dall-E are classic examples of generative AI platforms. If a user copies and pastes meeting minutes into ChatGPT and asks it to write up a brief, neatly organized summary of the meeting, ChatGPT will generate a summary based on those notes.
In healthcare, generative AI tools can be used to increase administrative efficiency by speeding up tasks such as summarizing doctors’ notes or translating text into a patient’s native language. Other proposed uses for generative AI in healthcare include developing new drug candidates based on data on existing molecular structures and improving the quality of medical images like MRI scans. It’s important to note, however, that generative AI tools tend to “hallucinate,” making up false information that may appear to come out of nowhere. For this reason, human oversight is key whenever AI is used in the healthcare system.
When it comes to AI in translation, for example, AI can play a helpful role in producing fast and mostly accurate translations of medical documents and patient records—it can even be used for real-time communication between LEP patients and their caretakers. But it’s not a replacement for translators and interpreters altogether. Humans who are familiar with both the source and target language need to review and vet AI translations to ensure that the AI output doesn’t contain any inaccuracies or hallucinations.
How Are Hospitals & Healthcare Providers Using AI?
Here are a few examples of hospitals and healthcare providers that are using AI to enhance patient care:
1. Partners Healthcare: When Partners Healthcare’s COVID-19 hotline was overwhelmed with callers during the pandemic, the organization implemented an AI chatbot to act as a screening tool. In addition to screening callers for COVID-19 symptoms, the chatbot could also answer most questions about the virus and advise patients on whether they should visit an urgent care or emergency room.
2. Johns Hopkins Medicine: Doctors at Johns Hopkins Medicine can use AI to draft responses to patient messages in the patient portal, editing the generative AI response as needed. The team at Johns Hopkins is also working on ways to summarize charts.
3. Mount Sinai: Since 2013, doctors at Mount Sinai have been using algorithms to identify patients in their system who are more likely to get sicker. By using this information, they’re able to more effectively treat them before their conditions worsen significantly.
Benefits of AI in Healthcare Applications
Using AI in translation can have several benefits for healthcare providers, such as:
- Increased efficiency: Generative AI tools can produce translations in a matter of seconds. This means that doctors and other caretakers can act faster, as they don’t have to wait as long to receive a translation.
- Better scalability: AI tools can translate huge amounts of content within a relatively short time period. This means that translation projects can be scaled as needed without much trouble.
- Improved accessibility of multilingual translations: Due to better efficiency and scalability, multilingual translations have become more accessible than ever.
Risks of AI in the Healthcare Industry
Despite the benefits, using only AI in translation carries several risks that healthcare providers must be aware of:
- Legal compliance concerns: Healthcare providers must comply with regulations surrounding patient data privacy, such as HIPAA. Not all generative AI platforms are compliant, so it’s important to make sure that you utilize tools that are specifically intended for use within a healthcare setting to ensure compliance with these laws.
- Accuracy concerns: As previously mentioned, generative AI tools have a tendency to hallucinate, meaning they might generate false information or something unrelated to the prompt. With AI in translation, for example, it’s not unheard of for the AI to generate a new sentence or paragraph that was not present in the source text. This means AI translations could be inaccurate, potentially damaging the quality of care a patient receives.
- Quality concerns for different languages: AI models vary in quality from language to language. AI models tend to produce fairly fluent text in English, but not so much in lower-resource languages. A recent study found significant discrepancies in how an AI tool answered questions about infectious diseases in English versus Arabic, suggesting that AI is not a reliable tool for every language.
Best Practices for Implementing AI in Healthcare Translation
Although the benefits of AI may seem tempting, the risks of using AI on its own outweigh the advantages. Healthcare providers can still take advantage of the benefits of AI in translation by following these best practices:
- Work with language experts to incorporate human oversight: The best way to verify that a translation is accurate is to work with a human translator who specializes in healthcare translation and is familiar with both the source language and the target language. Human translators can edit the text to make sure the translation does not contain any hallucinations or culturally insensitive language.
- Implement quality assurance measures: Quality assurance measures like back-translation (translating AI output from the target language back into the source language) can help you gauge the quality of a translation as well.
- Train AI models with domain-specific datasets: Using datasets that focus on terminology and jargon that are commonly used in a healthcare setting can help render more accurate translations.
- Ensure compliance with relevant regulations: AI tools meant for a general audience, such as ChatGPT, typically aren’t HIPAA compliant—look into generative AI solutions that are specifically tailored to a healthcare setting, and check to see how they ensure compliance with relevant regulations.
Future Outlook of AI in Translation and Healthcare
As the technology behind AI improves, you can expect to see AI being used more and more in the healthcare setting. AI has especially great potential for automating translation and administrative tasks—it’s possible that this technology will eventually be used to automate the translation of electronic health records.
However, as AI becomes more and more widespread in the healthcare system, ethical concerns are bound to arise. Many concerns around the ethics of AI have already come up—for example, the idea that inaccuracies due to the use of AI could negatively impact a patient’s health. Looking toward the future, it will become important to address these concerns and figure out a way to balance the efficiency of AI with human factors like empathy and judgment.
Frequently Asked Questions About Generative AI in the Healthcare Industry
How is AI helping the healthcare industry?
AI is improving efficiency across healthcare by automating administrative tasks and supporting translation for patients with LEP. It can also be used to more efficiently analyze medical images and detect early stages of disease, among other proposed applications of the technology.
Is Generative AI safe for LEP patients?
Not on its own. While AI can speed up translation and improve access, human oversight is essential to ensure accuracy, cultural sensitivity, and compliance with healthcare regulations.
Can healthcare providers rely solely on AI for document translation?
No. AI-powered translations can be fast, but they often contain errors or omissions. Providers should always involve professional medical translators to review and verify AI-generated text.
What are reliable AI translation technologies that can be used in healthcare settings?
Healthcare providers should use AI tools specifically designed for medical contexts and compliant with regulations like HIPAA. These tools, paired with the expertise of human translators, ensure safer and more accurate outcomes.
Conclusion
AI can improve healthcare translation workflows by making things faster, more scalable, and more efficient. But as powerful as AI may be, it can’t replace the expertise, cultural awareness, and judgment of human translators. As with other kinds of translation technology, AI requires human oversight to ensure that it’s being used safely. The most effective approach is a hybrid one—leveraging AI to speed up translation while relying on human professionals to ensure accuracy, compliance, and patient-centered care.
For healthcare providers, the next step is clear: Embrace AI as a valuable tool, but implement it thoughtfully. By following best practices—such as incorporating human oversight, using domain-specific datasets, and ensuring regulatory compliance—hospitals and providers can deliver better, more inclusive care to patients of all backgrounds.
At Avantpage, we enable healthcare providers to bridge language gaps between care teams and their LEP patients using technology-forward translation and interpretation workflows with a human in the loop. If you’re looking for a language access solution, Avantpage has you covered. Contact us today at [email protected] or (530) 750-2040, or fill out this form for a free quote.
