It's extremely difficult to quickly obtain "reliable insights" from them. Open Evidence is a US-based AI paper navigation tool for medical professionals that directly addresses this challenge. In this article, I will introduce this tool that you will definitely want to use in clinical medicine.
What is Open Evidence and what are its strengths?
Open Evidence is an AI-based paper review support platform designed for healthcare professionals. It cites and generates answers to clinical questions. This may sound similar to the numerous AI paper search tools I have introduced on this blog. What sets this tool apart is its "partnerships with top-tier journals and medical institutions to ensure high-quality evidence." This reliability is supported by partnerships with world-renowned medical journals and medical institutions.
- JAMA (Journal of the American Medical Association)
- NEJM (New England Journal of Medicine)
- Mayo Clinic
These are institutions and journals well-known in the medical industry, but access is normally restricted because they are paywalled, and regular AI paper search tools cannot access their full-text content. However, Open Evidence generates answers based on papers from these top-tier journals. NEJM and JAMA reviews, including relatively new content, are comprehensively summarized and often useful in clinical practice. The fact that you can use all of this significantly increases the quality of the information compared to other tools.
Looking at the details, Open Evidence has a contract with the NEJM Group (New England Journal of Medicine) as of February 19, 2025, to provide full text and multimedia content from 1990 onwards (OpenEvidence). The JAMA Network has also been integrated into the AI platform, with a contract signed on June 5, 2025, for the adoption of full-text and multimedia content from 13 journals (Fierce Healthcare). Furthermore, Open Evidence has a collaborative relationship with the Mayo Clinic, and many actual physicians are involved in its development.
In addition to NEJM and JAMA, the tool frequently cites international (primarily US) guidelines, making it highly practical for clinical use. These features are, I believe, very valuable for clinical physicians.
Setup
To use it, you can install the app from the official website or on an iPhone, and then register as a user.
Official website ▼ https://www.openevidence.com/
In addition to your name, you are required to prove that you are a healthcare professional (HCP). Since it is basically a service for the US, there is no clear information on how it works in Japan. I sent an email to the official OpenEvidence site to inquire about this, but received no response at all. The privacy policy seems to state that usage outside the US should comply with each country's laws, but it seems that it is not officially supported.
How to Use It
- Enter a clinical question or keyword. Example: "Does SGLT2 inhibitor reduce heart failure hospitalization in elderly?"
- Relevant papers are automatically extracted, and the answer is organized and displayed with citations. Cited answer text References This part is the same as regular AI answer generation tools. However, a key feature is that the references are limited to reviews, guidelines, and highly cited papers, as mentioned above. In fact, clinical mega-journal reviews are often substantial and helpful, so if you have a point of interest, diving deeper into the review can be useful for your clinical practice.
This tool can also be asked questions in Japanese. However, like other tools, you need to be careful about translation errors that can occur due to language differences.
After advancing to graduate school and getting more research time, I strongly feel that the time I have to investigate simple questions from the clinical field has significantly decreased. While being busy with research, it's hard to find the time to delve into something that piques my interest on the spot. Furthermore, I often struggle with not having enough time to keep up with the latest information and treatment updates in areas outside my specialty, or in other departments.
In this situation, a tool that can provide reliable information quickly and allow you to check the source paper is a great help in both daily learning and clinical decision-making.
Comparison with other AI paper search tools
The number of AI-powered paper search tools has been rapidly increasing in recent years, but there are only a limited number of tools that can be used directly in a clinical setting. Here, I'll briefly summarize the differences between Open Evidence and other AI paper search tools.
General search tools like Perplexity, ChatGPT, and Gemini
Search tools based on LLMs are convenient because they allow you to search for guidelines and review papers using natural language. However, they have the following limitations:
- Access to paid papers is limited, so summaries and citations are restricted.
- It is difficult to filter searches to "only find guidelines" or "only latest reviews."
- There is a lack of transparency and consistency in sources. In contrast, Open Evidence's clear strength is its high accuracy in extracting reviews and guidelines because it is based on the full text of major journals like NEJM and JAMA.
Paper search tools like Scispace, Elicit, Paperguide, and Answerthis
These tools specialize in paper summarization and question-answering functions, making them useful for research purposes. However, they still have issues when used clinically:
- Access to paid papers is often unavailable, meaning information from those papers cannot be obtained.
- They may cite papers based on basic research or animal experiments, which are not directly applicable to clinical practice. Open Evidence's big advantage is that its filtering is clinically appropriate, as it is designed for practical use in the clinical setting.
Drawbacks
While I've praised it a lot, it is still a generative tool based on a large language model. You need to be careful about hallucinations, and it is still necessary to check the source citations. Since many large medical institutions have contracts with NEJM and JAMA, it's necessary to check the content if you can access the original articles. In my experience using it, I sometimes saw answers that were slightly different from clinical common sense (and which weren't supported by the original source).
Furthermore, perhaps because it is easy to reference review articles, some of the very latest information can be missing. For example, when I asked about tenecteplase, a thrombolytic agent used for cerebral infarction, it answered that the drug was not yet approved, even though it had already been approved by the FDA in the United States (as of August 2025). This could be a side effect of its focus on a limited set of references.
Also, while it has the advantage of breaking through paywalls, if you don't have a contract with the journal, you can't verify the original source, which can be problematic. It's not advisable to use it without being able to verify the information, so in such cases, other tools might be better.
Additionally, since it is a tool for finding evidence for clinical questions, it does not include features for managing or meticulously reading papers. For managing and reading papers, the other paper search tools mentioned above would have an advantage.
Summary
Open Evidence is a US-based AI paper navigation tool for healthcare professionals. Its main feature is its ability to search and summarize full-text content of paywalled papers that are normally access-restricted, thanks to its partnerships with top medical journals like NEJM and JAMA, and institutions like the Mayo Clinic.
Users can input a clinical question in natural language, and the tool automatically extracts relevant papers and presents a summary and clinical implications in an organized way. The abundance of included guidelines and review papers makes it particularly practical for clinical use, which is a major difference from other AI paper search tools.
On the other hand, since it is based on a large language model, the risk of hallucinations remains, and confirming the sources is essential. It is also not suitable for paper management or detailed reading, so other tools may be more appropriate for research purposes.
Overall, it is a powerful support tool for physicians and healthcare professionals who want to obtain highly reliable clinical evidence quickly. While it won't completely replace existing AI search tools, it can be said to have pioneered a new field in the clinical domain.
コメント