In what we believe is the first comperative study of its kind, NoBarrier AI's medical interpreter has demonstrated superior performance over traditional phone interpretation services, marking a significant milestone in healthcare communication technology.

The Study Design

Our research team developed a comprehensive benchmark of 91 medical sentences that mirror typical provider-patient conversations.  The text reflected real-world conversation between an English speaking doctor and a Spanish speaking patient. They included common patient expressions like describing symptoms ("I've been having sharp chest pain at night"), medical history questions from providers ("Have you had any previous surgeries?"), treatment explanations ("You'll need to take this medication twice daily with food"), and diagnostic information exchange.

We engaged two established global medical interpretation agencies, connecting with their phone interpreters just as healthcare providers do daily. The same sentences were processed through NoBarrier AI's interpretation system. To ensure unbiased evaluation, all interpretations were blindly assessed by two external certified medical interpreters. The results were analysed by Dr. Michal Schuster, a medical interpreting expert.

Key Findings

Overall Accuracy

The results were remarkable. On a 7-point accuracy scale, with 7 being perfect translation and 1 being completely incorrect, NoBarrier AI achieved a 6.38 average score, significantly outperforming traditional agencies which scored 5.59 and 4.93 respectively. This represents a 25.3% higher accuracy rate for NoBarrier AI compared to the average of traditional services.

Error Analysis

The total error count revealed an even more striking difference. Across all 91 sentences tested in both directions:

• NoBarrier AI: 64 total errors
• Vendor 1: 135 errors
• Vendor 2: 187 errors

This means NoBarrier AI made less than half the errors of traditional services, a crucial difference in medical settings where accuracy is paramount.

Breaking Down Error Types

Our analysis identified four main categories of errors:

1. Information Omissions - leaving out important details from the original messagesome text
   • Traditional Agencies: 53-70 omissions
   • NoBarrier AI: Only 17 omissions
2. Medical Terminology Accuracy - incorrectly translating specific medical termssome text
   • Traditional Agencies: 18-23 errors
   • NoBarrier AI: Only 5 errors
3. Unnecessary Additions - adding information not present in the original messagesome text
   • Traditional Agencies: 32-55 instances
   • NoBarrier AI: 14 instances
4. Word Substitutions - using incorrect word choicessome text
   • Traditional Agencies: 32-39 instances
   • NoBarrier AI: 28 instances

The Language Direction Breakthrough

One of the most significant findings was in language directionality. Human interpreters typically perform better when working into their native language. However, NoBarrier AI demonstrated remarkable consistency across both translation directions:

• Spanish to English: 6.51 average score
• English to Spanish: 6.26 average score

In contrast, the human interpreters showed marked differences in performance depending on direction:

• First vendor: 6.2 (Spanish to English) vs. 4.99 (English to Spanish)
• Second vendor: 5.69 (Spanish to English) vs. 3.93 (English to Spanish)

This consistency is crucial for medical settings, where accurate interpretation is needed in both directions of the conversation. While human interpreters showed notable performance drops when working into their non-native direction, NoBarrier AI maintained nearly equal quality regardless of the translation direction.

‍

Why This Matters for Healthcare

The implications of these findings are profound for healthcare delivery. When interpreters miss information or mistranslate medical terms, it can lead to serious consequences:

• Delayed or incorrect diagnoses
• Misunderstood treatment plans
• Compromised patient consent
• Increased liability risks

Consider a patient saying, "I've been having sharp chest pain at night." If the timing ("at night") is omitted or the type of pain ("sharp") is mistranslated, it could significantly impact the diagnosis and treatment path.

NoBarrier AI's superior performance, particularly in reducing information omissions and medical terminology errors, directly addresses these critical challenges.

Looking Forward

This study represents a watershed moment in medical interpretation. While human interpreters have been the backbone of medical interpretation services, AI technology has now demonstrated its ability to not only match but exceed traditional interpretation quality.

The consistency, accuracy, and availability of AI interpretation could help address the concerning statistic that over 70% of US physicians report inconsistent use of interpretation services. By providing a more reliable, accurate, and accessible solution, we can work toward better health outcomes for limited English proficiency patients.

Our commitment at NoBarrier remains unwavering: to break down language barriers in healthcare through innovative technology while maintaining the highest standards of accuracy and reliability. This study confirms we're on the right path.

Want to Learn More?

• Interested in the detailed methodology of our study?
• Want to see more specific examples of medical interpretations?
• Curious about how we can implement this technology in your healthcare setting?

Contact our research team at info@nobarrier.ai for the complete study results and to discuss how No Barrier can help improve medical interpretation in your organization.

‍