Machine learning is fundamentally about creating a system from existing observations to predict outcomes for new, unseen data. Even simple linear regression, a common tool in medicine, is a machine learning model, making the concept more accessible to non-experts.
Open Evidence employs a two-step process that distinguishes it from general LLMs. First, its AI, trained by human subspecialists, identifies the most relevant scientific references for a query. Only then is an answer generated from this curated evidence, prioritizing source credibility over speed.
To maintain clinical reliability, Open Evidence restricts its knowledge base to a 'walled garden.' This includes PubMed abstracts (excluding predatory journals), guidelines, and full-text content from licensed publishers. This prevents the model from citing unreliable internet sources or 'hallucinating' references, building clinician trust.
The platform uses machine learning to combat the rapid obsolescence of medical guidelines. It systematically reviews every paragraph of ingested guidelines against new publications weekly, identifying and flagging specific sections that may be deprecated by more recent evidence, a critical function in fields like oncology.
By analyzing millions of queries from clinicians, Open Evidence identifies high-frequency topics. It then cross-references these with its literature database to pinpoint areas that are both clinically relevant and poorly supported by existing evidence, effectively mapping the frontier of medical knowledge and research opportunities.
After a new drug launches, clinicians query AI tools about unexpected patient reactions. Analyzing these queries at scale can serve as a hypothesis-generation engine for rare side effects not seen in smaller Phase 3 trials, effectively acting as an informal Phase 4 pharmacovigilance system.
Recognizing that different specialists have unique information needs, Open Evidence develops distinct AI models for each role. For example, a surgeon's query about a procedure can surface peer-reviewed surgical videos—a feature irrelevant to a medical oncologist—thus enhancing the tool's utility through deep specialization.