LLM, AI Improve Medical Report Clarity and Clinical Workflows

Large language models (LLMs) are being deployed to simplify radiology reports, enhancing patient comprehension and supporting clinical workflows in Mexico and across Latin America. Companies like Eden leverage AI and tele-diagnosis platforms to improve diagnostic accuracy, optimize operational efficiency, and integrate with local health systems while navigating COFEPRIS regulations and regional data protection requirements. The adoption of AI-driven tools addresses workforce constraints, rising patient volumes, and the need for more accessible, patient-centered care, with implications for hospitals, imaging providers, and health technology investors operating in Mexico.

Large language models (LLMs) are increasingly being applied to simplify medical reports, making radiology more accessible to patients while supporting clinical workflows. A recent meta-analysis published in The Lancet found that AI-driven simplification can reduce confusion, improve patient engagement, and streamline hospital operations, though clinical oversight remains essential. Companies like Eden are deploying these tools at scale across Latin America, demonstrating measurable impact on diagnostic accuracy and efficiency.

LLMs and Patient-Centered Care

A systematic review of LLM performance in medical report simplification, published in The Lancet, shows the dominance of OpenAI models, with 92% of studies using GPT-based systems. GPT-4 outperformed GPT-3.5 in sensitivity analysis, achieving a mean accuracy rating of 4.77 versus 4.09, improving clarity and comprehension. Other models evaluated included Google Gemini/Bard, Anthropic Claude, Microsoft Copilot, and open-source solutions such as Llama and Mistral.

Simplifying radiology reports addresses the longstanding "jargon barrier," where technical language often confuses patients, causing unnecessary anxiety and follow-up visits. Lowering reading levels to a school-age comprehension (ages 11–13) empowers patient agency, supports shared decision-making, and conveys empathy, with simplified reports scoring an average of 3.61 in perceived emotional sensitivity. The Lancet review emphasized that AI tools must be paired with clinician validation to avoid clinically significant errors, which occurred in roughly 0.9% of reports.

Julian Ríos, CEO and Founder, Eden, explained to MBN, “Eden was born from the mission to ensure no one suffers or dies due to delayed or incorrect diagnoses. Every decision, from hiring to product development, centers on this purpose.” Eden’s AI models, including its thorax AI for early-stage lung cancer detection, show the potential of combining LLMs with imaging-focused AI to improve diagnostic outcomes. The company reports a 50% reduction in transcription time for radiologists and serves over 15 million patients across 2,200 imaging departments in 18 countries, including leading hospitals like Médica Sur in Mexico and DASA in Brazil.

Beyond simplification, Eden integrates AI into clinical workflows to optimize operational efficiency. Its platform supports over 35,000 daily diagnoses and more than 80,000 physician and patient portal accesses per day, guiding treatment decisions from routine imaging to complex interventions such as pediatric oncology protocols or heart transplant planning. By controlling both algorithms and workflow, Eden ensures that AI not only produces accurate results but also translates into tangible improvements in patient care.

Adoption and Trust Challenges

While the potential benefits of AI-driven health simplification are clear, adoption is tempered by trust and accuracy concerns. Patients increasingly use AI for early symptom research, but surveys reveal caution: only 13% of respondents agreed that AI-written summaries are usually accurate, and 35% cited privacy concerns. In Mexico, AI has become the third most consulted source of health information in 2025, up from eighth in 2024, highlighting rapid uptake alongside persistent skepticism.

High-profile cases, such as Google removing AI-generated health summaries after accuracy concerns, underscore the risks. Summaries on liver blood tests, for example, lacked sufficient context, prompting removal to prevent potential patient harm. Experts warn that oversimplified AI responses can mislead patients and reinforce misconceptions, a phenomenon termed “AI pandering” by Erick Ponce Flores, President, ICAN. Emotional chatbots and mental health support applications illustrate both opportunities and limitations of AI in sensitive areas, highlighting the need for mechanisms that prioritize safety, accuracy, and clinical oversight.

Health Union’s 2025 Connected Health Experiences survey found that one-third of patients had used AI tools like ChatGPT or Google Gemini, primarily as an initial research step rather than a definitive source. About two-thirds of respondents relied on AI-generated summaries to orient themselves, but fewer than half verified the original sources. Age differences were notable: patients under 50 reported greater familiarity with digital health tools, yet even younger users expressed concerns over reliability and loss of human connection.

The World Economic Forum and Boston Consulting Group reported that trust in AI hinges on technical literacy, adaptable regulatory frameworks, and public-private collaboration. Flexible models, regulatory sandboxes, and lifecycle monitoring are essential to ensure that AI solutions remain safe and effective post-deployment. Without coordinated frameworks, fragmented approaches could exacerbate inequities between higher and lower-resourced regions, particularly in Latin America.

Scaling AI in Latin America

The global market for AI in clinical workflow is projected to grow from US$2.78 billion in 2025 to US$11.08 billion by 2030, reflecting a 31.9% compound annual growth rate. AI supports diagnostics, results interpretation, administrative tasks, and care coordination, particularly in inpatient and multi-specialty settings where high patient volumes and complex workflows demand efficiency gains. Eden’s deployment across 2,400 departments illustrates how integrated AI platforms can improve operational productivity while maintaining clinical oversight.

Eden’s approach emphasizes local adaptation. Its AI models are trained on Latin American radiology data to account for regional anatomical and physiological variations, addressing a key limitation of models trained exclusively on US or East Asian datasets. The company also integrates AI with tele-diagnosis and workflow optimization, enabling measurable improvements in both accuracy and throughput. Ríos explained, “Transforming workflows and culture requires resilience, field engineering, and close collaboration with medical teams. Scaling to 18 countries in four years illustrates both opportunity and challenge.”

Proposals to integrate AI into Mexico’s healthcare system aim to expand coverage, improve care quality, and strengthen data protection. Meanwhile, COFEPRIS’ new administration faces regulatory delays, intellectual property disputes, and modernization challenges, influencing how AI-driven solutions are evaluated and deployed. The combination of AI adoption, workflow integration, and regulatory adaptation will determine how quickly Latin American health systems can leverage these tools to meet rising patient volumes and workforce constraints.

Across the region, the integration of AI is also shaping patient-provider interactions. Patients benefit from faster comprehension of diagnostic results, but technology must supplement rather than replace human care. Hybrid models that combine AI insights with physician guidance are increasingly seen as the most effective approach, reinforcing trust while promoting efficiency.