35. Bridging Language Barriers in Plastic Surgery: A Systematic Review of AI as a Medical Interpreter

Abstract

BACKGROUND: Effective communication is critical in plastic surgery for quality care, better outcomes, and patient satisfaction. Limited English proficiency (LEP) is associated with shorter surgeries, lower satisfaction, and fewer follow-up visits. Surgeons may use their language skills or bilingual staff/family as interpreters when professional ones are unavailable, risking compromised informed consent and understanding. AI offers real-time language translation for patient-provider interactions. This review evaluates AI’s effectiveness as a medical interpreter in clinical settings, focusing on accuracy, usability, satisfaction, and feedback, giving insights on future directions in plastic surgery. METHODS: A systematic search was conducted on July 11, 2024, using CINAHL, IEEE Xplore, PubMed, Scopus, Web of Science, and Google Scholar. Studies needed to discuss AI as an interpreter for patient-clinician exchanges. Exclusions included review articles, correspondence, educational materials, non-peer-reviewed/retracted reports, non-English articles, publications before 2016, and sign language or patient education studies. Search terms for AI, language interpretation, and healthcare were adapted per database, following PRISMA guidelines. Two investigators independently screened, extracted, and synthesized results, with bias assessed using ROBINS-I, MMAT, or JBI Critical Appraisal Checklists. Discrepancies were resolved by a third reviewer. RESULTS: From 1,095 reports, 9 studies met the inclusion criteria, evaluating AI translation platforms like Google Translate, Microsoft Translator, Apple iTranslate, AwezaMed, Pocketalk W, and the Asynchronous Telepsychiatry (ATP) App. Studies were conducted in the United States, France, Switzerland, and South Africa, published between 2019 and 2024. AI translation showed promise for brief communications, such as postoperative visits, with accuracy ranging from 83-97.8% translating from English and 36-76% to English. Usability scores ranged from 76.7% to 96.7%, with patient satisfaction between 84-96.6% and healthcare professional satisfaction between 53.8-86.7%. However, clinicians expressed hesitancy due to concerns about respect, quality, reliability, and potential misunderstandings. Currently, AI is used as a last-resort option, to assist non-certified providers, for brief interactions, and alongside lay interpreters. CONCLUSION: AI’s integration into plastic surgery can enhance LEP patient communication, particularly in brief postoperative discussions. However, further advancements are needed before its use in complex clinical conversations. Limitations include a narrow range of languages, lack of bidirectional translation assessment, and evolving tools. While AI translation tools demonstrate efficiency and high patient satisfaction, a hybrid approach combining AI’s rapid translation capabilities with human translators’ nuanced understanding is critical to ensuring high standards of care in plastic surgery.

Ähnliche Arbeiten