The NEJM Catalyst study examined how Spanish speaking patients perceive different interpretation services modalities during a postoperative consultation.
The study showcases remote video interpretation through Jeenie (RVI) and an AI medical interpreter powered by No Barrier.
The central finding is contextual preference.
Patients did not ask for replacement. They asked for the right modality at the right moment.
When Do Patients Prefer Human Interpreters?
Patients preferred human interpreters for emotionally complex or high stakes conversations.
“If I’m scared or it’s something delicate, I prefer a person” (Participant J.R.).
For healthcare leaders, this reinforces that emotionally sensitive encounters remain a human anchored domain.
When Do Patients Prefer an AI Medical Interpreter?
Patients highlighted speed and constant availability as defining advantages of the AI-based interpreter.
“I can use it wherever I am, at any time” (Participant A.L.).
In emergency departments and perioperative workflows, instant access reduces delay and throughput friction.
An AI medical interpreter changes response time from scheduled to immediate.
Does Privacy Influence AI Adoption?
Yes. Privacy emerged as a clear driver.
“With AI I feel like no one else is listening” (Participant D.B.).
For some patients, removing a third party from the interaction increased comfort during sensitive discussions.
This insight has implications for disclosure accuracy and patient trust.
Are There Concerns About Dialect Accuracy?
Yes. Dialect variation was explicitly raised.
“AI may not understand some words we use in my country” (Participant M.P.).
“Sometimes Spanish is different between countries — the words aren’t always the same. I’m not sure if the artificial intelligence would understand those differences” (Participant J.L.).
Spanish vocabulary differs across regions. In medical settings, word choice can affect symptom reporting, consent clarity and discharge comprehension.
No Barrier, the AI-based interpreter evaluated in the study supports five Spanish variations to reflect regional differences: Spanish from Mexico, Cuba, the Dominican Republic, Puerto Rico and US Spanish. Dialect modeling remains an ongoing refinement priority. Dialect is a precision issue.
How Does Interpreter Shortage Factor Into Deployment?
Interpreter scarcity is structural in many hospitals.
“I can use it wherever I am, at any time” (Participant A.L.) reflects more than convenience. It reflects availability.
An AI medical interpreter scales access. Human interpreters remain essential for complex emotional encounters.
The hybrid model addresses both capacity and sensitivity.
Executive Summary
The NEJM Catalyst study shows:
- Human interpreters are preferred for emotionally delicate conversations
- AI medical interpreters are preferred for speed and privacy
- Dialect precision remains a critical quality factor
- Hybrid deployment addresses interpreter shortages
The future of healthcare interpretation is not substitution. It is intelligent integration.
Importantly, this was not a live care deployment. No Barrier and Jeenie were not tested in real clinical workflows. Instead, the study used controlled simulation to isolate modality preference.
For healthcare executives, that distinction matters. The findings inform perception, trust and adoption readiness.
Special recognition to Dr. Gezzer Ortega MD MPH Physician Scientist at Brigham and Women’s Hospital and one of the study’s authors for advancing evidence in this space. We also had the opportunity to host Dr. Gezzer Ortega for an open discussion on the Care Culture Talks podcast focused on culturally competent care and the operational realities of language access in modern health systems.
Key Takeaways for CMOs
- Interpretation modality should align with clinical urgency
- Dialect precision affects comprehension and safety
- Hybrid models protect empathy while improving flows
- Instant AI access reduces operational drag across surgical workflows
Source: NJEM Catalyst Study, Patient Preferences and an Implementation Framework
.svg){kind=small}
%201.png)