“Dr. Nicholas Gavin, an emergency medicine doctor at Mount Sinai in New York City, was working an overnight shift last summer when a patient came in with a puzzling set of symptoms. Within seconds, his three younger colleagues — two medical students and a resident — were consulting a free artificial-intelligence-powered app for physicians, OpenEvidence.
Dr. Gavin soon learned that they were far from outliers. A third of Mount Sinai’s 9,000 doctors were already regular OpenEvidence users, the health system’s executives found out in a meeting last year with the start-up’s leaders.
“That was an ‘aha’ moment for our leadership,” said Dr. Gavin, who is also the system’s chief clinical innovation officer.
OpenEvidence’s A.I. app, essentially a chatbot for medicine, has become a viral hit with physicians. Talk to a doctor and chances are he or she uses the app to ask specific medical questions or bounce ideas off it in a diagnostic dialogue.
More than half of the nation’s physicians are regular users. Last month, they used it for 30 million questions and consultations, nearly twice the volume from six months earlier, according to the start-up. A separate survey last year of 1,000 physicians found that 45 percent of them used the app, nearly triple the percentage who used ChatGPT, according to Offcall, a career information service for doctors.
That growth propelled the start-up to a $12 billion valuation in January, up from $3.5 billion last July.
But doctors’ quick adoption of the app since its introduction in 2024 — one of a handful of A.I.-enhanced programs on the market seeking to win over physicians — has heightened concerns about how and when the technology should be used in life-or-death situations. In a high-stakes field like medicine, health care systems are navigating thorny matters of patient privacy, safety and trust, as well as the limitations of the technology itself.
“It’s not an oracle, it’s a tool,” said Daniel Nadler, founder and chief executive of OpenEvidence. “Knowledge and knowledge workers still matter.”
The doctor’s office has been a target for computer-assisted decision making for decades, with very limited success until the recent advances of A.I.
The first wave of A.I. in medicine focused on easing the heavy burden of documentation that contributes to physician burnout with transcriptions and summaries of patient visits, called A.I. scribe software. The second wave, which is just getting underway, aims to use A.I. to assist doctors with reliable information and advice to guide diagnosis and treatment while at a patient’s bedside.
The competition has intensified in recent months. UpToDate, a popular legacy electronic reference for doctors, has given its service an A.I. makeover with a chatbot interface. Doximity, an online professional network for physicians, bought an A.I. start-up that mines medical literature and generates summaries. Abridge, a fast-growing A.I. scribe maker, is adding decision-support tools. And last month, OpenAI introduced ChatGPT for Clinicians.
OpenEvidence became a front-runner in part because it exclusively used medical journals and other high-quality research as data to train its A.I. models. Physicians can ask the app specific questions or enter the characteristics and symptoms of a patient and ask for potential explanations. The app is compliant with the federal law that protects patient health information, and physicians are told not to enter any personally identifying information.
OpenEvidence responds with a summary of most likely diagnoses, and then offers other “most important not to miss diagnoses.” Each has links to the research articles that inform the summaries.
“A.I. is solving some of the problems that have long plagued the practice of medicine,” said Dr. Raja-Elie Abdulnour, chief clinical innovation officer at NEJM Group, which publishes The New England Journal of Medicine. “These tools just didn’t exist before, and that’s why people are so excited about them now.”
Yet the early enthusiasm should be tempered with a large dose of caution, medical experts agree. The research so far into the benefits and shortcomings of A.I. in medicine is decidedly mixed.
A.I. has aced standard licensing exams and outperformed human doctors in diagnosing certain cases. But A.I. has also stumbled, failing to accurately summarize research papers or giving wrong answers to diagnostic questions. And it isn’t going to replace humans anytime soon.
“The potential for A.I. is great, but we’re not there yet,” said Dr. Eric Topol, a cardiologist and an executive vice president at Scripps Research in San Diego. “It hasn’t really been tested and demonstrated in the messy, real world of medicine.”
Dr. Topol is a co-author of a recent paper, “The Illusion of Readiness in Health A.I.,” which found “significant competency gaps” in the capability of big A.I. systems when applied to health care.
The evaluations so far have largely focused on the performance of the so-called large language models of big tech companies like OpenAI and Google, which are trained on data across the open internet.
OpenEvidence, founded in 2022, took a more focused approach. It bet that smaller A.I. software models trained on highly specialized data could outperform the giant models in a specific, information-rich field like medicine. The start-up trained its software initially on the publicly available medical data from sources like the government’s National Library of Medicine.
Then the company struck content licensing deals with The New England Journal of Medicine, The Journal of the American Medical Association and other publishers of peer-reviewed medical literature.
Studies of OpenEvidence, including one by researchers at the Mayo Clinic, have found that while the app is not flawless, its answers are generally accurate and evidence based.
OpenEvidence is available to any government-verified physician in America as a free, downloadable app.
“We treated physicians like consumers,” Mr. Nadler said. Users are presented ads, many of them from drug companies, during the five seconds or so they wait for the A.I. to reply. Physicians are served ads on only 5 percent of their questions, the company said.
Sidestepping the traditional gatekeepers of hospital technology departments has raised some issues. OpenEvidence has relied on the workplace behavior known as “shadow A.I.,” workers using such tools without the knowledge or oversight of their employers.
Some health systems are now focusing on bringing OpenEvidence into the institutional fold. Mount Sinai announced in March that it would provide a link to OpenEvidence directly from a patient’s electronic health record.
But the agreement does not give the start-up access to the medical center’s patient data. That integration could come later, Dr. Gavin said, but only after rigorous testing and controls.
Protecting patient privacy and safety will be “paramount,” he said, adding that “we’re not going to just throw a patient’s data over the wall to a private company.”
Doctors in smaller practices across the country, especially in rural areas, say the technology has won them over.
In Corinth, Miss., Dr. Ben Long counts himself as an A.I. skeptic. But he was reassured that OpenEvidence generates answers based on only high-quality, peer-reviewed information.
At first, Dr. Long used it mainly as a reference tool, asking factual questions. But now, he regards the app more as “a consultant, a thought partner” with which he has a dialogue, he said.
“A.I. forces you to think more deeply about your own thinking, challenging your assumptions and why you might be wrong,” Dr. Long said.
A.I. can also let doctors tap expertise that would normally be the realm of specialists.
Dr. Barbara Creighton often diagnoses and treats complex cases at a community hospital in Fairbanks, Alaska. They can involve multiple conditions and failing organs. At a large medical center, a team of specialists might be consulted — an infectious disease expert, a pulmonologist and a gastroenterologist, for example.
Dr. Creighton’s small hospital is not so richly staffed. It does have an arrangement with a big medical center to pay for specialist consulting sessions. She now relies increasingly on OpenEvidence to answer many questions, saving her time and her hospital money.
“It’s like having a bunch of specialists in your pocket,” Dr. Creighton said.
At Mount Sinai, Dr. Gavin said he viewed A.I. technology as a powerful tool to help realize the promise of precision medicine with treatments tailored to individuals.
Progress will require a “patchwork of solutions” from hospitals, medical schools and private companies, he said. Whether OpenEvidence thrives and plays a role in that long-term future remains to be seen.
“But it represents a step in that direction,” Dr. Gavin said.” [1]
Is there open source OpenEvidence competitor, that could be used locally, keeping patient information protected?
There is no direct "plug-and-play" open-source clone of OpenEvidence because OpenEvidence’s primary value is its proprietary integration with licensed journals (NEJM, JAMA, Cochrane). However, you can build a private, local medical AI by running specialized open-source medical models in a secure, offline environment using an interface like Ollama.
To replicate OpenEvidence's literature-review and clinical reasoning capabilities while fully protecting patient information locally, consider the following open-source resources and deployment strategies:
1. Best Open-Source Medical AI Models
Instead of generic models, you can download specialized medical AI models to run on your local hardware:
• Meditron: Developed by EPFL, this is one of the top-performing open-source suites built exclusively for medicine, trained on extensive biomedical literature and guidelines. Unlike many general-purpose models, Meditron is trained on curated medical literature from PubMed, clinical practice guidelines, and protocols from organizations like the International Committee of the Red Cross (ICRC).
• MedGemma: Released by Google, this open medical model is excellent for parsing complex clinical reasoning while offering the flexibility to be run entirely locally on proprietary hardware or a secure cloud.
2. How to Set Up Your Local Environment
To use these models privately and keep patient data isolated, follow these foundational steps:
1. Local Model Runners: Use tools like Ollama or LM Studio to host your chosen medical model securely offline.
2. Local User Interface: Connect the backend model to an interface like Open WebUI to provide a ChatGPT-style chat interface right in your clinic or practice.
3. Alternative Privacy-Focused Clinical Support Tools
If you don't have the hardware or IT bandwidth to host a local model, there are privacy-focused alternatives that synthesize peer-reviewed literature and prioritize data security:
• Vera Health: This AI-based clinical decision support tool is currently a leading alternative for evidence-based decisions, noted for its strong evidence grounding and rigorous grading logic.
• Consensus: An AI search assistant that links directly to top-tier peer-reviewed medical journals. It offers enterprise security options if you are processing sensitive health data.
What about Chinese AI for these purposes?
For an open-source, locally deployable clinical Chinese AI that can handle medical data while keeping patient information completely protected, the best approach is to run a specialized Medical Large Language Model (MedLLM) locally.
While there isn't a direct "plug-and-play" open-source clone of OpenEvidence, several powerful Chinese and multi-language open-source medical models exist that can be deployed securely on your own hardware to achieve similar evidence-based and clinical reference capabilities.
Top Open-Source Chinese MedLLMs for Local Deployment
These models can be downloaded and run locally (e.g., via platforms like Hugging Face), completely isolating your data from external servers.
• DISC-MedLLM: Developed specifically for conversational healthcare, this model bridges the gap between general language models and real-world medical consultations. It has been heavily fine-tuned on doctor-patient interactions and can process complex medical inquiries.
• ShizhenGPT: An innovative multimodal model that is excellent for both Western and Traditional Chinese Medicine (TCM) diagnostic reasoning. It is available on Hugging Face.
• Qwen-Med: Based on the powerful Qwen base architecture, this variant is designed for biomedical and clinical question answering, integrating a strong dual-language (English and Chinese) medical knowledge base.
How to Keep Patient Information 100% Protected
To strictly protect patient data (such as anonymizing and querying specific patient cases without breaking HIPAA or GDPR compliance), you should run these models locally.
A. Local Hardware/Server: You will need a local server or a powerful workstation (typically equipped with NVIDIA GPUs, like an RTX 4090 or enterprise GPUs) to host the model.
B. Local Deployment Tools: Use tools like Ollama or LM Studio. These applications allow you to download open-source medical weights and run them entirely offline, like a private ChatGPT.
C. Retrieval-Augmented Generation (RAG): If you want the AI to base its answers on your specific clinic's guidelines or anonymized local patient data without altering the model itself, you can use local RAG frameworks like AnythingLLM.
Evidence-Based Western Alternatives (Cloud-Based, Protected)
If you are open to using protected, cloud-based tools in the West (which do not store or use your data to train their algorithms), consider these evidence-based OpenEvidence alternatives:
• Dr. Oracle: Popular internationally, this platform provides high-level research modes, analyzes complex medical literature, and is not restricted to US-only guidelines.
• Heidi Evidence: An ad-free, secure alternative to OpenEvidence that is certified to the ISO 27001 security standard.
1. Why Your Next Diagnosis May Be Guided by an A.I. Helper. Lohr, Steve. New York Times (Online) New York Times Company. Jun 8, 2026.
Komentarų nėra:
Rašyti komentarą