Before language access was standard practice, limited-English proficient (LEP) patients were found to have longer hospital stays than English-speaking peers who underwent similar procedures and surgeries. Today, language access supports timelier care and more equivalent outcomes. As hospitals continue to drive improvements for LEP patients, can emerging tools such as AI translation be used to bridge communication differences?
AI exhibits significant potential to impact the changing landscape of patient care. In some applications, such as in medical and pharmaceutical research, AI has already led to positive outcomes. However, this technology also has its limitations and remains far from supplanting the role of physicians and nurses in diagnostics and treatment.
The same is true for certified medical interpreters. When evaluating AI tools for language access in a hospital setting, administrators have to weigh the ethics and the possibility of risk. There are a few use cases where AI can offer a scaled solution for language access, particularly when translation services are needed. This does not mean AI is equivalent to the capabilities of human-powered language services, especially for sensitive, confidential conversations with providers.
According to our assessment, certified medical interpreters for hospitals will remain irreplaceable for the foreseeable future. Here’s why.
Healthcare AI: Opportunities and Limitations
AI in healthcare presents opportunities to help patients on a larger scale. At the same time, these tools are in their early stages and may introduce new variables that are not well-controlled. Where and when is this emerging tech appropriate for use in a patient-facing setting?
Healthcare administrators are already using AI to streamline repetitive, high-volume tasks, such as billing, insurance, scheduling, and compliance assessments, so long as tools meet HIPAA and privacy requirements. Public health researchers have leveraged AI successfully to screen patient data for early detection or risk factors of certain cancers. Clinicians are increasingly finding value in AI tools that surface data and trends that might go overlooked, which can be useful when making diagnostic and treatment decisions. In drug research and development, AI also has shown promise for identifying new therapy targets and flagging possible adverse interactions.
With these exciting advancements in mind, it is important to consider AI’s limitations. Not enough randomized clinical trials have been conducted to validate the applications of this emerging technology in clinical practice. Further, independent research has found widespread biases towards ethnic and racial minorities in training data, algorithms, and the design of AI tools, which would create more harm than good if replicated on a mass scale. This reality underscores the importance of caution when introducing AI tools into healthcare. Clinicians must remain the primary decision-makers about diagnosis and treatment.
Certified Medical Interpreters vs. AI in Healthcare: The Use Case Matters
These insights about healthcare AI can be applied to the context of language access in patient care. Technology has aided in the expansion of language access in hospitals, primarily through virtual remote interpreting (VRI), where LEP patients can see and listen to their interpreter through a device, and the interpreter can see and hear both the provider and patient. By connecting in minutes rather than waiting for an on-site interpreter to become available, more patients and languages are served. AI-powered, human-supervised translation can also be leveraged effectively to benefit patients, especially for scaled, repeated projects, such as translating consent forms, patient summaries, and other important documents.
When communicating with LEP patients, however, the role of certified medical interpreters cannot be replaced with AI tools. The precision, cultural competence, and sensitivity required of medical interpreting exceed AI’s capabilities. If providers have to rely on an imperfect AI tool for conversations with LEP patients, particularly considering AI cannot pick up on non-verbal cues, they become exposed to risk. Missing information from a patient’s history, medication list, or current symptoms may lead to misdiagnosis, adverse outcomes, exposure to malpractice or discrimination lawsuits, or even HIPAA violations if the technology is not compliant.
When integrating new technology into patient care, the use case matters. Just because AI can surface data about patient cases does not mean it should be relied on for diagnoses. Similarly, AI-powered voice interpreting tools may be able to facilitate superficial understanding in low-stakes situations, but that does not mean they are ready to supplant the nuance, capabilities, and judgment of trained interpreters. The risk of errors—and the impact of those errors—are simply too high. Your language access plan should identify clear boundaries for when the use of these tools is permitted or prohibited.
Hospital Risk Management: Evaluating AI Tools for Language Access
When creating a hospital language access plan, the role of risk management is to assess and mitigate liability, whether through compliance with non-discrimination laws or ensuring providers are able to communicate effectively with patients. For example, most language access plans already prohibit using patients’ family members as interpreters due to most providers’ inability to assess whether information is being relayed accurately. Also, many LEP patients may be hesitant to share private medical information with family members or friends, especially considering adults who may rely on their children to interpret for them.
Most AI tools will raise red flags with risk management, and for good reason. AI algorithms process large volumes of data to produce the deep insights the technology promises. If the platform is not HIPAA-compliant, ensuring that patient data is not processed outside of the hospital’s data environment, risk management cannot allow it.
Additionally, there is the issue of transparency. A significant challenge is the black-box nature of many AI algorithms. AI models are more likely to make an error, rather than identify they do not know something with certainty, further compounding the challenge. Healthcare providers and risk management professionals alike often struggle to understand how these systems reach their conclusions, leading to hesitancy in trusting and implementing AI-driven recommendations.
If you intend to incorporate AI tools into your language access plan, working with a HIPAA-informed partner can help drive success. Ideally, your partner will strike a balance with investments in emerging technology that are advancing hospital language access capabilities while remaining transparent about the limits of its applications to patient care. They should also be able to provide solutions that align with HIPAA requirements.
Evolving Your Hospital's Language Access Plan
The National Health Law Program analyzed medical malpractice claims involving language barriers that may have led to patient harm. Across 35 claims, insurance carriers paid a collective $2,289,000 in damages or settlements and $2,793,800 in legal fees. The group estimates language access issues comprise 2.5 percent of malpractice lawsuits. This cost far exceeds the expense of investing in language access services, especially considering the efficiencies of video remote interpreting (VRI).
Your language access plan likely already accounts for the risks associated with malpractice lawsuits. Does it also provide strategies to address the risks introduced by AI?
PGLS can help evolve your hospital’s language access plan. Through a consultative approach, our hands-on team can help you identify care gaps in the LEP patient journey, assess and mitigate risk, and clarify technology’s role in language access. You can count on our world-class team of translators and interpreters to handle healthcare-related content with sensitivity, accuracy, and cultural awareness. In over 200 languages and regional varieties, we facilitate patient-provider communication in their preferred language remotely, over-the-phone, over video, and on-site so they can understand and make informed decisions about their care.
Contact us today to learn more, and download our free eBook to create a language access plan for your hospital.